CA2099175A1 - Device for high efficiency encoding and decoding of picture signals and recording medium - Google Patents

Abstract

ABSTRACT
To realize a high picture quality with a small information volume, to reduce the hardware scale and to reduce the capacity of the frame buffer of a decoder.
A limitation mode decision circuit 34 adaptively changes over a mode of inhibiting interframe predictive coding over the entire macro-blocks in each slice to a mode of inhibiting interfield predictive coding in a frame being encoded over the entire macro-blocks in one slice. As for a B-frame, prediction from its odd field to its even field is inhibited, while prediction from an odd field, such as an Io field, of a reference frame of forward prediction ia also inhibited.

Description

- ~- 2 ~ 7 ~ C ~ ci o TIII,E OF THE IN~r101~: Device for ~-ligh E~ficiency Encoding and Decoding of Picture Signals and Recording Medium BACKGROUND OF THE INV~TION

1. Field o~ the Invention This invention relates to a technique for high efficiency encoding Oe image sign~ls by orthogonal transform9 a recording medium on which data encoded by ~he high efficiency enccding tec~nique is rscorded, and a technique ~or decoding the encoded data.

2. Related Art As a system ~or high ef~iciency encoding o~ image signals, the Moving Picture Experts Group (MPF~) has proposed a draft standardization which prescribcs a high e~ficiency encoding arrangement ~or picture signals for digital storage media.
The storage media contemplated in the proposed arrangement are those having a continuous transfer rate o not higher than 1.5 M bit/sec, such as a compQct disc (CD), digital audio tape re~order (DAT) or a hard disc. Aocording to the draft standardization, the storage m~dium may be directly connected to a decoder or it may be connected thereto via a transmission medium such as a oomputer bus, local ar~a network (LQN) or tel ~ ication link. It is cont~mplated by the dra~t : standardiza~ion to implement special ~wlctions, such as random ac~essing, high-speed playback or reverse playback, in addition to the usual ~orward playback.
In the arrangement proposed by MPEG ~or high e~iciency encoding of picture signals, redundancy along the time axis is lowered by encoding the di~ference between pictures9 and subsequently redundancy along the spatial axes is lowered by employing discrete cosine transform (Dcr) and variable length encoding.
With respect to redundancy along the time axis, in moving pictures, a picture under consideration, that is a picture at a given point in time, bears a strong resemblance to ~emporally previous and temporally posterior pictures. Consequently, if the di~erence between the picture now to be encoded and the picture temporally preccding it is taken and transmitted, as ., ,, . : ... . . . .

`~ 2Q~17,j shown in Fig.44, it becomes possible to lower the redundancy along time axis to diminish the volume of information to be transmitted. The pieture encoded in this manner is termed a forwardly predictive-coded picture, P picture or a P frame.
Similarly, if the difference is taken between the picture now to be encoded on the one hand and the picture temporally preceding it and the picture temporally su¢ceeding it or interp3lated pictures produced ~rom the precedin~ and succeeding pictures, and a smaller one o~ the resulting diferences is transmitted, it becomes possible to lower ~urther the redundancy along time axis to diminish the volume of in~ormation to be transmitted. The picture encoded in this manner is termed a bidirectionally predictive-cod~d picture, B picture or a B frame.
The pictures shown at I, P and B in Fig.44 represent an intra-coded picture, I picture or I frame, as later explained, as well as the above-mentioned P picture and B picture, respectively.
The respective predicted pictures (referred to as predictive-coded pictures) are produced by so-called m wement cQmpensation.
As an example of movemcnt compensation, a block consisting o~ e.g. 16 * 16 pixels, which is ~ormae~l oP Pour unit blocks each of e.g. 8 * 8 pixels, is prepared. The 16 * 16 pixel block is termed a macro block. The macro-block of the temporally previous picture in the vicinity of a given macro-block of a currsnt picture exhibiting the least difference from the macro-block of the current picture is found by searching, ~d a difference therebe~ween is taken ~o enable reduction of the data to be transmitted. For the P-picture9 a pieture produced by taking the difference between the picture now to be enooded and the predioted picture following movement ~o~pensation or a picture produced without taking the difference between the picture to be encoded and the predicted picture following movement compensation, whichever exhibits less data volume, is selected and encoded from one 16 * 16 pixel macro-block to another.
However~ in the above case, more data needs to be transmitted for a picture portion which has just emerged at the back of an object which-has been moved. For a B-picture, decoded temporally previous or temporally posterior pictures following movement compensation, an interpolated picture produced by adding these pictures, or a picture to be encoded, whichever is the least in data vol~me, is encoded.

" 2~1r~
With respect to redundancy along spatial axes, the differences of the picture data are not directly transmitted, but are discrete oQsine transformed from one 8 * 8 pixel block to another. The discrete cosine transform (DCT~ expresses a picture not on the pixel level, but as a function of which Or the frequency components o a cosine function are contained in a picture and the ,~mounts oP the ~requency compQnents contained in that picture. For example, data contained in a 8 * 8 pixel unit block are transformed by two-dimensional DCT into a block of 8 * 8 coefficients of the co~ponents of the cosine function.
It is known that picture signals of a natural scene taken by a - television camera tends to be smooth si~nals. In such case, the data volume may be efPiciently diminished by discrete cosine trans~orming the picture signals.
Thus, in the ease o~ smooth signals~ such as picture signals of the natural scene, larger values are concentrated about a certain coefficient. If the coefficient is quantized, ~he 8 * 8 coefficient block bec~mes substantially equal to zero, while larger coefficients are le~t. For transmitting data o~
the 8 * 8 ~oefficient block, non-zerc coefPicients and a zero run indi~ating how many Os are present in front of the c~efficient a~e grouped into a set in the sequence Or a so-called zig-zag scan, which set is transmitted (Huffman code) or diminishing the amount of information needed to be transmitted. The image is re-cons~ructed in the reverse sequen~e at the decoding side.
The da~a structure handled by the above-described encoding arrangement is shown in Fig.45 and is c~mprised of a block layer, a macro-block layer, a slice layer, a picture layer, a group o~
pictures (GOP) layer and a video sequence layer, looking from bottom to top in Fig.45. The various layers are explained beginning ~ram the lowermost layer.
Rererring to the block layer, sach block of the block layer is made up o~ 8 * 8 luminance or color dif~erence pixels, that is 8 neighboring luminance pixels in 8 lines or 8 neighboring ~olor diferenoe pixels in 8 lines. The above-mentioned DCr is performed for each o~ these 8 * 8 blocks.
In the macro-block layer, each macro-~lock of the layer is made up o ~our leet, right, upper and lower nsighboring luminance blocks YO, Yl, Y2 and Y3, and color di~Perence blocks (unit color difPerence blocks) Cr and Cb which are at the same 2 0 9 ~ ~ 7 ~
position in the picture as luminance blocks. These blocks are transmitted in the sequence of YO, Yl, Y2, Y3, Cr and Cb. Which of the blocks is used as a reference picture for taking differences or i~ no di~ference needs to be transmitted is determined by the enc~ding system fro~ one macro-block to another.
- Ihe slice layer is made up of on~ or plural macro-blocks which are contiguous in the picture scanning sequence. At a head part of the slice, the motion vector and DC component in the picture are reset. The ~irst macro~block has data indicating a position in the picture so that restoration is possible on error occurren~e. Consequently~ the length o~ a slice or the starting position of the slice is arbitrary and may be changed depending on the error states of transmission rinks.
In the picture layer, the pictures are ~ach made up o~ at leas~ one and pre~erably plural slices. Depending on ~he encoding system, the pictures are classified into the above-mentioned intra-encoded pictures, I pictures or I-frames, the ~orwardly predictive-coded pictures, P-pictures or P-~ramcs, bidirectionally predictive-coded pictures~ B-pictures or B-frames, and DC intra-coded pictures, or DC coded (D) pictures.
In the intra-coded picture or I-picture, only the information conta;ned with the particular picture is employed. In other words~ the picture may be re-~onstituted on decoding only by the in~ormation o~ the I-picture. In effect, the picture is directly discrete cosine trans~ormed and encoded without taking a difference between it and any other picture. Although the encoding system is generally low in e~iciency, I-pictures may be inserted at arbitrary positions to enable random access and high-speed playback.
In the ~orwardly predictive-coded picture or P-picture, the I-picture or P-picture which is temporally previous and has already been decoded is employed as a reference picture for di~ference taking. Either the difference between the present picture and the movement-compensation reference picture or the present picture itsel~, without taking a direrence (intra-coded picture), whichever is more eficient? is selected ~or encoding, and this selection is made from one macro-block to another.
In the bidirectionally predictive-coded picture or B-picture, three reference pictures, namely the temporally preceding alre~dy decoded I picture or P-picture and interpolated pictures 2 0 9 v ~. 7 ~
produced ~rom both o~ these pictures, are employed as the reference pictures. Encoding Or one o~ the three di~erences ~ollowing the movement cGmpensation or intra encoding, ~hichever is most e~ficient, is selec~ed from one macro-block to another.
The DC intra~coded picture is ihe intra-coded encoded picture constitut~d only by thè DC ~oefficients of DCr and c~nnot exist in the same sequence as the other three pictures (i.e. the I, P
or B pictures).
The group-of-picture (G~DP) layer is made ~p o~ one or plural I-pictures and zero or plural non-I pictures. If the input s~quence into the enc~der is lI, 2B, 3B, 4P~5B, 6B, 7I, 8B, 9B, 10I, 11B, 12B, 13P, 14B, 15B, 16P*17B, 18B, 19I, 20Bl 21B and 22P, the output sequence of the encoder9 that is the input sequence to the decoder is lI, 4P, 2B, 3B*7I, 5B, 6B, 10I, 8B, 9B, 13P, 11B, 12B, 16P, 14B, 15B*19I, 17B, 18B, 2P, 20B and 21B.
The reason such change in the sequence is made in the encoder is that, for enoGding or decoding the B-picture, it is necessary that the reference picture9 that is the temporally pos~erior I-picture or P-picture, be present at the time the B-picture is to be encoded or decoded. The distance between the I-pictures or the distance between the I-picture and B-pi¢ture, is arbitrary.
Besides, the distance between the I-pictures or the P-pictures may naturally be c ~ ged within the GOP layer. The boundary between tha GOP layers is represented by *. The I-picture, P-picture and the B pic~ure are represented by I, P and B~
resp~ctively.
The video sequence layer is constituted by one or plural GOP
layers having the same picture sîze, picture rate, etc.
When a normalized moving picture is to be transmitted by the high efficiency enc~ding arrangement o~ MPEG, a picture obtained by compressing data within the picture is transmitted, and then a difference from the same picture processed with movement compensation is transmitted.
When a field, for example, is processed as a picture, verti~al positions become different between two fields, so that di~ference data need to be transmitted for transmitting e.g. a still picture.
For processing a frame as a picture, a picture deformed in a c~nb shape needs to be processed as long as a moving portion in the frame is concerned. For example, in Fig. 3, if a moving object CA, such as a car, is present ahead o a stationary 2~ J~7~
background, since movement ~ccurs betwccn Pields of a ~rame, the picture ~or the portion of the moving object bccomcs zig-zag-shaped in contour as indicated at KS.
For processing a pieture in which a stationary portion and a moving porion exist together, there is produced a picture por$ion in the picture which is low in compression efficiency, regardless of whether the ~ield is pr~cessed as a picture or the frame is processed as a picture.

O~JECT AND SUMMARY OF THE INVE~ON

It is an object o the present invention to provide a high efficiency encoding technique for encoding picture signals whereby field or frame processing of a field-based video picture may be carried out efficiently regardless o~ whether the picture contains a subjcct that undergoes little motion or abundant motion.
It is also an object of the present invention to provide a decoding technique which is the counterpart to the encoding t~chnique; to pr wide a recording mcdium on which the en~oded data is recorded.
In accordanoe with one aspect of the present invention, a picture signal encoding technique is provided for encoding interlace scan pictures by transforming pictue data representing the interlace scan picture by adaptively selecting ~rame-based or field~based ~ .
In accordance with another aspect o~ the present invention, a picture signal decoding technique is provided for decoding in~erlace scan pictures that have been encoded ny inverse transforming the encoded interlace scan pictures by adaptively select;ng ~rame-based or field-based IDCT to produce picture data.
In accordance with a further aspect o~ the present invention, high efPiciency encoding apparatus is provided for encoding picture signals as a ~unc~ion Oe a macro-bl~ck whic~ consist oP
a two-dimensional array of plural pixels, includes a motion detector for detecting Prame motion vectors between frames on the macro-block basis and for detecting field motion vectors between fields on the rnacro-block basis, a ~irst mode selector for selecting on the basis o~ the macro-block, either a ~rame 2 ~ 9 v ~ 7 ~
prediction mode for carrying out frame motion compensation or a field prediction mode for carrying out field motion compensation, a se~ond mode selector ~ar selecting on the basis o~ the macro-block either a ~ramc processing mode ror transforming a block o~
frame or a field processing mode for transforming a block of field data, a predictive encoder for cncoding input picture data by using the selected frame or field motion c~mpensation based on the frame or field motion vectors and the frame or field prediction mode to produce first encoded data, and a transform encoder ~or en~oding the ~irst encoded data by using frame or ~ield orthqgonal transformation as a function of the ~rame or field processing mode, In accordance with yet another aspect of the present invention, a high efficiency decoding apparatus is provided for decoding picture signals~ this apparatus comprising an inversc variable length deccder for d ~ ing encoded data to reproduce motion vector information, prediction mode inrormation (which indicates the blo¢k division ~or motion compensation), pr~cessing I~Dde information on the basis of whether a frame in a macro-block or a field in the macro-block is more efficient) ~nd encoded pic~ure data, an inverse transformer for decoding ~he encoded picture da~a by using the inverse orthogonal transform as a function o~ the processing mode in~ormation to produce first decoded picture data~ a prediction decoder ~or decoding the ~irst decoded picture data by using mot]ion compensation based on the mo~ion vector and the prediction mcde.
In accordance with a still urther aspect o~ the present invention, a recording m~dium has record~d thereon an encoded bit stream comprising, en~oded picture data, prediction mode data, motion vector da~a and processing mode data.
The above, and other objects, features and advant~ges o~ the invention, will be apparen~ in the following detailed description o~ an illustratiYe embodiment ther~of which is to be read in connection with the aocompanying drawings forming a part hereof, and wherein corresponding parts and components are identified by the same reference numerals in the several views of the drawings.

2 ~ 9 v 1 7 ~
BRIEF DESCRIPTION OF THE ~RAWINGS

Fig.1 is a block diagram showing a schematic arrange~ent of a high ef~iciency encoding apparatus of a furter em~odiment.
Fig.2 shows an example o~ macro-block.
Fig~3 shows an example of macro-block ~or a ~rame processing mode.
Fig.4 shows an example of macro-b]ock for a ~ield processing mode.
Fig.5 is a block diagram showing an arrangement of a high efficiency encoding device for picture signals of the second embodiment.
Fig.6 shows the manner of encoding by the ensoding devi~es of the first and second embodiments.
Fig.7 shows a unit block ~or DCT o~ the frame pr~cessing mcde/~isld processing mcde for a typi~al concrete format of a digital VTR.
Fig.8 shows the manner of motion prediction in Fig.7.
Fig.9 shows a modification of Fig.7.
Fig.10 shows the manner of motion prediction in the mcdification of ~ig.9.
Fig.11 shows a unit block Por DCT of the frame processing mcde/field processing mode for another typical concrete format of a digital VTR.
- Fig.12 shows another modification of Fig.7.
Fig.13 shows a set oP macro-b]ocks.
Fig.14 shows the manner of processing in accordance with the ~rame processing mode with Fig.13, Fig.15 shows the manner o processing in aocordance with the field processing mode with Fig.13.
- Fig.16 shows a modi~ication (for forward prediction) of extension bit addition in the encoding device oP second embodiment.
Fig.17 is a block diagram showing an arrangement of a decoder which is a counterpart of the encoding devices of first ~nd second embsdiments.
Fig.18 shows a picture of an odd cycle.
~ ig.l~ shows a picture of an even cycle.
Fig.20 is a block diagram showing a schematic arrangement of a high eP~iciency encoding device ~or pic~ure signals o~ the third embodiment.

~ O ~ V ~ ~ ~
Fig.21 is a block diagram showing a schematic arrang~ment of a high efficiency encoding device fGr picture si~nals of the fourth embodiment.
Fig.22 is a block diagram sowing a schematic arrangement of a modification of the high efficiency enc~ding device for picture signals of third embcdiment.
~ ig.23. is a flow chart for explaining a modification 1 of processing by limitation mode selecting means in the high efficiency encoding device for picture signals of third embodiment.
Fig.24 shows motion vector from an odd field to an even field.
Fig.25 is a flow chart for explaining second modification of processing by limitation mode selecting mcans in the high efficiency encoding device for picture signals of third embodiment.
Fig.2B is a flow chart for explaining third modifi~ation Or processing by limitation mode selecting means in the high efficiency encoding device for picture signals of third embcdiment.
Fig.27 is a flow chart for explaining fourth modification of processing by limitation mode selecting means in the high efficiency encoding device for picture .signals of third embodiment.
Fig.28 is a flow chart for explaining fi~th mcdification oP
processing by limitation mode selecting means in the high efficiency encoding device for picture signals of third embodiment.
Fig.29 shows motion prediction for the se~ond limitation mode and frame pro~essing mode.
Fig.30 shows motion prediction for the second limitation mode and field proeessing mcde.
Fig.31 shows motion predic~ion for the first limi~ation mode.
Fig.32 is a block diagram showing a s~hematic arrangement of an encoding device (modieication~ o~ the second e ~ iment.
Fig.33 is a block diagram showing an arrangement of a third dec~ding device.
Fig.34 is a block diagram showing a schematic arrangement of a high eficiency encoding device for picture signals oP the fifth embodiment.
Fig.35 shows the code decoding and display sequence by the Pifth encoding device.
Fig.36 shows the code decoding and display sequence by the second (or third) encoding device.

2~ 17~
Fig.37 shows the manner o~ motion prediction in the fifth encoding device.
Fig.38 is a block diagr~m showing an arrangement o~ the fi~th decoding device.
Fig.39 shows the code dec~ding and display sequence by the fifth decoding device.
- Fig.40 is a block diagram showing a schematic arrangement of a high efficiency encoding device for picture signals of the sixth embodilnentO
Fig.41 shows the code decoding and display sequence by the sixth decoding d~vice.
Fig.42 shows the manner of motion prediction in the sixth encoding device.
Fig.43 is a block diagram showing an arrangement of the sixth decoding device.
Fig.44 shows several predic~ion pic~ures.
Fig.45 shows a data structure.
Fig.46 shows a picture having a moving object.

- I o -2 Q ~
DESCRIPTION OF PREF~RRED EMBODIMENTS

By referring to the drawings, preferred embodiments o~ the present invention will be explained in detail.

*** FIRST EMBoDIMENr ***
Fig.1 shows an ~mbcdiment 1 o~ the high efficiency encoding device for picture signals i which eneod;ng is performed with a macro-block, which is a two-dimensional array of pixels srnaller than a screen, consisting e.g. of 16 * 16 pixels of input picture data arrayed in a s~atial sequence o~ input picture datag as a unit. The encoding device incudes a group o~ frame memories 10 for storing plural frames, each consisting of plural unit blocks of 16 * 16 pixels, as original pictures, and a fr~me movement detection circuit 22 and a ~ield movement dete~tion unit 21, as movement detection means, for detecting the sum of the differences of absolute values of the pixels and the movement veotors between the frames or between the ~ieldsO Each ~ield is composed of odd-numbered or even-numbered scanning lines of the frame pixels and is divided on the basis of the above-mentioned macro-bloc~ as a unit. The encoding device also includes a m~vement prediction mode decision circuit 23 and a s~.lector 249 as first mode selecting means~ for deciding, based on the output information of the movement detection ~ans, which of the frame predictive mnde or movement ccmpensation bascd on the frame in the macro-block as a unit or the field predictive mode for movement ccmpensation based on the ield in the macro-block as a ~nit and selecting the prediction mode with a higher efficiency. The encoding device also includes a block-forming mode decision circuit 25, as second mode selecting means, ~or deciding9 ~ased on the output information ~rom the ~ovement detection means and the first mode selecting means, which of the frame processing mode of forming blocks for ef~ectuating orthogonal transform based on the frame in the macro-blo~k as a unit or the field processing ~ode of forming blocks for e~fectuating orthogonal transform based on the rield in the macro-block as a unit is more ef~icient for orthogonal transform and selecting the m~de with a higher e~ficiency. The encoding device also includes an address generator 11, as a address generating means, for recognizing whether the cycle is an odd-7 ~-3 numbered cycle of scanning the odd-numbered field or an even-numbered cycle of scanning the even-numbered field in the interlaced scanning of an encQding operation ~or a frame and for controlling the group of ~rame nK~nories ~or out~utting macro-blocks forn ~ at the odd-numbered cycles for the selected block-forming mode. The encoding device also includes a group o~ frame memories 20 ~itted with a movement comp~nsator, as moven~nt compensating n~ns, for receiving the b]ock-forming mode information as selected by the second n~e selecting means and the movement prediction n~e information as selected by the first mode means and executing prediction movement-con~nsated fields or ~ramesO
The main stream o~ picture data to be encoded in Example 1 is explained with reference to the arrangement shown in Fig.l.
In Fig.l, digital picture signals are supplied ~o input terminal 1 so as to be stored in the group of frame memories 10.
Data of the 16 16 unit macro~blocks are r~ad from the group o~
frame m~mories 10, under control of address generator 11 as later explained, and transmitted to a di~erence detection unit 12.
Movement-compensated picture data from the group of frame memories 2D are also supplied to the difference detection unit 12 where the differences in the pic~ure data are detected.
An output of the differenee detection unit 12 is supplied to a DCr circuit 13 for effectuating orthogonal transform (DCT~.
DC~ coef~icient data rom the DCr circu;t 13 are transmitted to a quantizer 14. Quantized data from quantizer 14 are transmitted to a variable length encoding circuit 16 for e~fectuating a variable length ~ncoding, such as HuP~man encoding or run-length encoding, and outputted at output terminal 2 via buffer 16 as enccded data.
The quantized data fron the quantizer 14 are supplied to the frame memory group 20 fitted with the movement ccmpensator via dequantizer 17 for effectuating dequ~ntization which is the reverse of the quantization effectuated in the quantizer 14, an inverse nCT circuit 18 for efPectuating an inverse ~CT operation which is the reverse of the DCT operation per~ormed by DCT
circuit 13, and an additive node 19. The additive node 19 adds an output of the inverse DCT circuit 18 and an output of the frame memory group ~0 fitted with the movement compensator.
Meanwhile, signals for inhibiting overflow of the buffer 16 are 2 ~ 9 v ~ 7 ~
fed back from buffer 16 to quantizer l4.
On the other hand, picture data outputted from frame memory group 10 from one macro-block to another are tr~nsmitted to a frame movement detection circuit 22 and a ~ield rnovement detection clrcuit 21.
The frame movement detection circuit 22 detects and the motion vectors between frames and the sums of the absolute values of the differences between pixe]s from one macro-block to another and outputs these data, that is data o~ motion vec~ors between frames FMMV and the sums of the absolute values of the differences FMAD. The field movement detection circuit 21 detects the sums of the absolute values of the differences of the pixels from one macro-block to another to output these data, that is data of motion vectors between fields FDMV and the sums of the absolu$e values of ~he differences FDAD. The motion vector data F~V/FDMY of these motion vector detection circuits 21, 22 are transmitted to selector 24, while the da~a of the sum of the absolute values of the differences FMAD/FDAD are transmitted to the movement prediction mode decision circuit 23.
The movement prediction mode dccision circuit 23 decides, based on the data o~ the sums o~ the absolute values o~ the differences FMAD from frame mov~ment detection circuit 22 and the data of the sums o~ the absolute va]ues of the differences FDAD
from field move~ent detec~ion circ~it 21, as to whether the movement prediction is to be made on the ~rame-by-frame basis or on the field-by-field basis at the time of the m w ement prediction at the frame memory group fitted with the movement compensator 20, and outpu~s data indicating a processing mode which is more advantageous or efficient. SpeciPically, i~ the di~erence betw~en the sums o~ the absolute values o~ the differences FMAD and FDAD is found to be larger than a predetermined threshold Tl (FMAD - FDAD > Tl) by the movement prediction mode decision unit 23, data indicating that the field-by-ield movement prediction is more efficient (data MPFD for the field processing mcde for movement prediction~ is outputted ~rom circuit 23. Cbnversely, if the difference between the sums of the absolute values of the differences FMAD and FDAD is found to be smaller than or equal to a predetermined threshold Tl (FMAD -FDAD @Tl), data indicating that the frame-by-framc movement prediction is more efficient (data MPFM for the frame processing 2 ~ v .J J_ ! 3 mode for movement prediction) is outputted from circuit 23. The outputted movement prediction mcde data ~FM/MPFD is transmitted to the frame memory group ~0 fitted wit movement compcnsator which then effectuates frame-by-frame or field-by-field movement co~pensation. The movement prediction mode data MPFM/~PFD are also transmitted to sclector 24.
The selector 24 selects, responsive to the motion prediction mode data MPFM/MPFD ~rom motion prediction mode decision circuit 23, the ~rame-to-frame motion vector data FMMV supplied from frame motion detection circuit 22 or the field-to-field motion vector data FD~V supplied from field motion detection circuit 21.
That is, if the motion prediction m~de data is the da~a MPFD
indicating the field prediction mode da~a MPFD~ selector 24 selects and outputs the motion vector data ~DMV ~rom field motion detection circuit 21, whereas, if the motion prediction mode data is the data MPFM indicating the fr ~ e pre~iction mode data MPFM, selector 24 selects and outputs the motion vector data FMMV from rc~me motion detection circuit 22. The motion vector data FMMV/FDMV, as selected by selector 24, is trcm smitted to the block-formin~ mode decision circuit 25.
The block-forming mode decision circuit 25 is also supplied with output data from the field memory ~roup 10 and ~he processing mode data MP~M/MPFD ~rom motion prediction mcde decision circuit 23. The block~orming mode decision circuit 25 receives the motion prediction mode data MPFM/MPFD and the motion vector data FMMV/FDMY and formulates a differential picture using pictures from the ~rc~me memory group 10. The circuit 25 also selects, based on the difference picture, t~e block-forming mcde most suited to the picture processed by the DCT circuit 13.
For the I-picture or I-frame, data of the picture of the frame memory group 10 (original picture) is e~nploye~ in place o the above-mentioned differential picture.
It is now assumed that e.g. the macro-block of the differential picture is a macro-block shown for example in Fig.2.
In ~he case of the I-picture, the macro-block is the macro-block of the original picture. In Fig.2, ~dd-numbered lines ol, o2, o3, ..., oN, where N indicates 16 in the case of a macro-block, are indicated by solid lines, while even-numbered lines el, e2, e,3, ..., eN, where N indicates 16 in the case of a macro-block, are indicated by broken lines. The pixcls of the even-numbered lines are indicated as e(i, j), while those of the odd-numbered lines are indicated as o(i, j). In the differential picture or original picture~ that is the picture of the I-picture, as shown in Fig.2, the difference EFD of the field-by-field differential plcture may be represPnted by the equation 1, whereas the difference EFM o~ the frame-by-frame di~ferential picture may be represented by the equation 2.
[Equation 1] ~6 ~FD~ o~ o(i~ e(~j) e~

[Equation 2~ 16 l~
&j~ o~,J) ~ o(i~l,f) I (2) If a difference between the ~rame-by-frame differenoe EFM
and the field-by-field diference EFD, as found by the equations 1 and 2, respectively, is ~ound to be ]arger than a certain threshold T2 (EFM - EFD ~ T2), the block-rorming mode decision circuit 2S outputs data indicating that the DCr operation by the DCT circuit 13 be per~ormed on the ield-by-field basis, that is :~
data ~FD for the field-by-field operation mode for the block-~orming operation. Conversely~ if it is found that the difference between the differences E~M and EFD is smaller than or e~ual to the threshold T2 ~E~M - EFD @T2), the block-forming mode decision circuit outputs data indicating that the DCT :~operation by the DCT circuit 13 be perf'ormed on the frame-by-frame basis, that is data MDFM ~or the frame-by-frame operation mode ~or the block-~orming operation. The block-~orming mode data MDFM/MDFD is transmitted to the address generator 11 and to the frame memory group ~itted with movem~nt compensator 20.
Besides, the motion vector data (F~/F~), block-forming mode data (MDFM/MDFD) and the predictive mode data (MPFM~MPFD) are ~ransmitted to the variable leng~h encoding circuit 15.
The address generator 11 controls the frame memory group 10 to output the picture data stored therein in the ~orm of macro-blocks in accordance with the ~CT processing mode data MDFM/MDFD.
That is, i~ the bl~ck-~orming mode data is the data MDFM
indicating the ~rame-by-frame DCT operation, address generator 11 controls the ~rame memory group to output macro-blocks in which even-numbered and odd-numbered fields are scanned alternately with one another. Thus the unit macro-block transmitted to DC~ circuit 13 is constituted by the alternate even-numbered and odd-numbered fields. Conversely, if the block-, , ,~ .
.: . ' ';

. .
, . .

forming mode data is the data MDFD indicating the field-by-~ield DCT operation, address generator 11 controls the frame m~nory group to output a macro-block in whi~h the even-numbered field is separated from the odd-numberod field, as shown in Fig.4.
Thus the unit ~acro-block, transmitted to DCT circuit 13, is constituted by the ~dd-num~ered and the even-numbered fields separated from one another. However, the DCr circuit 13 per~orms DCT operation on the 8 * 8 pixel unit macro-block basis, as described previously. In Figs.3 and 4, the odd-numbered and even-nwnbered lines are indicated by solid and broken lines, respectively.
On the o~her hand, the predictive mode data MPFM/MPFD from ~he movement predictive mode decision circuit 23, the processing mode data MDFM/MDFD from DCT mode decision circuit 25 and the motion vector data FMMV/FDMY, as selected by selector 24, are also transmitted to the ~rame memory group fitted with motion compensator 20. Thus the frame memory group fitted with motion compensator 20 is not cnly responsive to predictive mode data MPFM/MPFD for movement prediction and to the bl~ck-forming mode data MDFM/MDFD for DCT, but also effectuates movemcnt compensation.

*** SECOND EMBODI~ENT ***
Fig.5 shows an embodiment 2 of the second high-ef~iciency encoding device of the present invention. In Fig.5, the blocks denoted by ~he s~me numerals as those in Fig.1 are operated in the same manner as those shown in Fig.1. Consequently, only the blocks bearing difPere~t numerals are explained.
That is, the high ef~iciency encoding device shown in Fig.5 include,s, besides the blocks bearing the same numerals as those o~ the high efficiency encoding device shown in Fig.1, a mcde decision circuit 43 and a selector 24, as processing n~de selecting n~ans, Por deciding, based on the output information ~rom the movement detection n~ans, which of the movement c~mpensation being oP the frame-by-f'rame predictive n~e and the block forming for orthogonal transform being of the field-by-field operating n~de or the movement compensation being of the Pield-by-field predictive l~ode and the block ~orming for orthogonal transrorm being of the fran~-by-frame operating n~de is n~re efPicient and selecting the more efPicient nKdes, and an .

2 0 ~ ~ 1 7 ~
address generator 31, as address generating means, for controlling the ~r~me memory group to recognize if the cycle in the interlaced scannin~ of each frame :for encoding is an odd-numbered cycle or an even-numbered cycle and ~o sequentially output odd-numbered fields for one macro-block in an amount corresponding to one frame at ~ach odd-numbered cycle only ir the mode of the m~de decision circuit ~3 is ~or ~ield prediction and field processing, and subsequently to output the even-numbered rields for one macro-block at each even-numbered cycle in an amount corresponding to one frame.
Meanwhile, the emkodiment 2 is directed to an encoding device for not separating the block-forming m~de from the motion compensating mode. Although the same block diagram as that o the embodiment 1 suffices, the embodiment 2 differs fundamcntally ~rom the embodiment 1 in the above-described operation of the address generator.
Meanwhile, the mode decision circuit 43 shown in Fig.5 (Embodiment 2) decides, ~ased on data of the sums of the absolute values of the differences FMAD ~rom the frame movement detection circuit 22 and data of the sums of the absolute values of the di~erences FDAD ~rom the ~ield motion fletection circuit 21, whether the movement prediction is to be performed on the ~rame-by-~rame basis or on the field-by field basis at the time of the m~vement prediction at the frame memory group fitted with movement compensator 20 as later explained, and formulates a differential picture, using the results of decision (prediction mode data ~PFM/MPFD o em~odiment 1)9 motion vector FMMV/FDMV
from the motion detection circuits 21, 22 and the picture from ~he frame memory group 10, ~or deciding, based on the differential picture, the mode ~or block-~orming most suitable to the picture processed with DCT by the DCT circuit 13. That is, the l~ode decision ¢ircuit 43 decides which of the motion prediction being of the frame prediction mode and the block-forming mode being oP the ~rame processing mode PDFM or the motion prediction being of the ~ield prediction mode and the block-~orming mode being of the field processing mode PDFD is more efficient. In other words, the mode decision circuit 43 has the combined ~unctions o~ the pr~diction mode decision circuit 23 and the block-forming mode decision circuit 25.
Meanwhile, the mode decision may be achieved in the same way ~9~ rtPj as the decision of the motion prediction mode and the blcck-forming mode in the ~mbodiment 1.
On the other hand, address generator 31 controls the erame m~mory group 10 to output the picture data stored therein in the form of macro-blocks in accordance with the aforementioned mode data PDFM/PDFD. That is, if the block-~orming mode data is the data PDFM indicating the frame-by-framc encoding operation, address generator 11 controls the framc memory group 10 to output macro-blocks in which sc3nning of even-m1mbered and odd-numbered ~ines is alternated with one ano~her. rrhus the uni~ macro-block transmitted to DCT circuit 13 is ~onstitut~d by the alternate even-numbered and odd-numbered ~ields. Cbnversely, if the block-forming mode data is the data PDFD indicating the field-by-field encoding operation, address generator 31 controls the frame memory group to output odd-numbered ~ields for the macro-block sequentially for one frame at the odd-numbered cycles and subsequently output even-numbered fields for the macro-block sequentially for one frame at the even-n~nbered cycles. Thus the unit macro-block, transmitted to DCT circuit 13 at the odd-numbered cycles, is constituted solely by the odd-numbered fields during the odd-numbered cycles and solely by the even-n ~ ered fields during the even-num~ered cycles. ~owever, ~he DCT circuit 13 effectuates DCT on the basis o~ the 8 * 8 pixel unit macro-blocks, as described previously.
That is, since it is possible with the above-described hi~h efficiency encoding devi~e for informat:ion signals, according to the ab w e-described e ~ iments 1 and 2., to make a changeover between the frame prediction mode ~nd the field prediction mode in the movement prediction ~nd between the fra~e processing mode and the field processing mode in the block-~orming for DCT9 on the macro-block basis, the encoding may be achieved most efficiently on the macro-block ~asis.
Speci~ically, the motion prediction and DCT as described hereinbelow is ef~ectuated by ~he en~oding device of the embcdiments 1 or 2 depending on the formats of e.g. the so-called di~ital VTR.
In Figs.6, 8 and 10, the ~ields making up a framc o~ ~he I-Prame or I-picture are indicated as Io field (an odd-numbered ~ield of the I-frame) and Ie ~ield (even-numbered field Or the I frame3, the ~ields making up the P-frame or P-picture are 2 ~ 'J ~ ~
indicated as Po field (an odd-numbered ~ield of the P-frame) and ~e fi~ld (even numbered field o~ thc P-frame), and the fields making up the B-frame or B-picture are indicated as Bo -~ield (an odd-numbered fie]d of thc B-fran~) and Bc ~ield (even-numbered field of the B-frame).
Meanwhile, the frame processing mode for bl~ck formation according to the embodiments 1 and 2 is that in which the odd-numbered ~ields and the even-numbered fields are combined to form the macro-block which is used as a proccssing unit. That is, the frame processing mode is that of forming a macro-block for each frame. On the other hand, the field processing mode for block formation is that in which the odd-numbered fields and the evPn-numbered fields each form macro-blocks which are used as processing units. That is, the field processing mode is that of forming a macro-block for each field. Consequently, with the I-frame, for example, the frame processing mode is changed w er fram the frame processing mode to the field processing mode or ViCP versa for each macro-block.
Besides, with the high e~iciency enco~ing device according to the ~mbodiments 1 and 2, each frame is divided, as long as the encoding operation is concerned, into Gdd-numbered cycles ~nd even-numbered cycles, corresponding to the periods o~ s ~ ~ning of odd-numbered cycles and even-numbered cycles of the interlaced scanning, respoctively.
Meanwhile, when handling the so-call~ed 4:2:0 component digital VTR format with the em~odim~nt 1, each macro-block is made up of luminance blocks YO, Yl, Y2 and Y3, each composed of odd-numbered ~ields ana even-numbered fields, and color di~ference bl~cks ~bO and Crl, each composed of odd-numbered fields, i~ the block forming mode is the frame processing mode, and DCT is performed based on the above-mentioned unit blocks of the macro-block, as shown in Fig.7. Cbnversely, if the block forming mode is the ~ield processing mode, each macro-block MB
is made up of luminance blocks Y020, Y130 composed of the odd-numbered ~ields, luminance blocks Y02e, Y13e each cQmposed of the even-numbered fields and color difference blocks CbO, Crl each compo,sed o~ odd-nwnbered ~ields9 and DCT is performed based on the above-mentioned unit blocks of the macro-block.
As for the motion prediction for the embodiment of Fig.7, motion prediction MCP between the I-frame and the P-frame is ~9~
possible for the frame predictive mode, as shown in Fig.8. On the other hand, motion prediction MCoPo between the lo field and the Po field, motion prediction MCoPe between the lo field and the Pe field and motion prediction ~K~ePe between the Ie field and the Pe Pield ~ e possible for the field prediction mode. That is, in Fig.8, motion prediction and b]ock formation may be present independently ~or the ~rame pr~diction/processing mode and for the ~ield prediction/processing mode, suoh that a motion vector may be found for the frame prediction mode and two motion vectors may be found ~or the field prediction mode.
Consequently, iP, in the above-descri~ed embodiment 1, the block-~orming mode of the I-frame is the frame processing n~de, the Io fields and the Ie fields are combined for the odd-numbered cycles to form the macro-block. During the odd-numbered cycles, DCT, quantization and variable length encoding are performed from one macro-block to another. Conversely, no data is transmitted ~uring the even-numbered cycles for the present mode. Meanwhile, DCT is perormed for the above-mentioned 8 * 8 unit blocks.
If the block-~orming mode is the field processing mode, the macro-block is constitutedl by an Io field and an Ie field, separated from each other, for the odd-numbered cycle, and DCT, quantization and variable length encoding are performed from one macro-block to another. Conversely, no dlata is transmitted during the even-numbered cycles, as shown in Fig.7. Meanwhile, DCT is per~or ~ ~or the above-mentioned 8 * 8 unit blocks.
For the P-frame, the following operation is performed. I, for example, the block formation mode of the P-frame is the frame processing mode and the motion prediction mode is the frame prediction m~de, the frame-to-~rame motion vector MVP is detected during the cdd-numbered cycles, using the forward picture (I-frame pic~ure) as the reference picture. The macro-block, composed of alternate Io and Ie ~ields, is used as a prediction picture, and the dif~eren~e thereof from the original picture is encoded. Conversely, no data is transmitted during the even-numbered cycles for the present mode.
If the blo~k-forming mode for the P frame is the frame processing mode, with ~he motion predictiorl being of the field prediction mode, the motion vector MVoPo between the Io field and the Po field, the motion vector MVePo between the le field and the Po ~ield, the motion vector MVoPe between the lo field and 2 ~ 7 ~
the Pe field and the motion vector MYePe between the Ie field and the Pe field are detected, using the To ~ield and the Pe field (or the Po ~ield and the Pe field) as the reference pictures, ~or the odd-numbered cycles. The one Or thc pr~diction for the od~-numbered cycles, the prediction for the even-numbered cycles and bo~h predictions, such as a ~ean value of the prediction ~or the odd-numbered field and the od-number~d ~ield, which will give the least value of the predicted error from the current frame P, is selected, and the di~ference from the original picture is found, using the above-mentioned macro-block, combined from the Io and Ie ~ields, as the reference p;cture. Cbnversely, no data is transmitted for the even-numbered cycles of ~he current mode.
If the bl~ck-~orming ~ode for the P-frame is the ~ield processing mode, with the motion prediction being of the ~rame prediction mcde~ the fr~ne-to-~rame moti~n vector MVP is detected for the odd-numbered cyclesi using the l-frame picture or the P-frame picture as a reference picture, and the difference from the original picture (which is a macro-block ~onstltuted by Po an Pe fields) is encoded, using the above-mentioned macro-bl~ck, constituted by the Io and Ie fields separated from one another, as a pr~diction picture. Conversely, no data is transmitted for the even-nwnbered cycles of the current mcde, in the same m~nner as above.
If the block-forming mode for the P :~rame is the field processing mode, with the mction prediction being of the field prediction mode~ the motion vector MYoPo between ~he Io field and the Po ~ield, the motion vector MYePo between the Ie fie]d and the Po field, the motion vector MYoPe between the ~o field and the Pe field ~nd the mction vector MVePe between the ~e field and the Pe field are detected, using the Io field and the Ie field (or the Po field and the Ps field) as the reference pictures, for the odd-nwnbered cycles. The one of the prediction ~or the odd-numbered cycles, the prediction ~or the even-nwnbered cycles and both predictions, such as a mean value of the prediction for the odd-numbered ~ield and that for the odd-numbered ~iel~, which will give the least value of the predicted error from the current ~r~me P, is selected, and the difference from the original picture, which is the macro-block combined from the Po and Pe fields, separated from each other, is encoded, using the macro-block, combined from the lo and Ie fields, separated ~rom each 2 ~ 7 `~
other, as the reference pieture. Conversely, no data is transmitted for the even-numbered cyclcs o~ the current ~de.
I~ the current fr ~ e is the B-frame, the following operation is performed.
If the block-forming mode for the B-frame is the frame processing mode, with the motion pr~d;ction being oP the frame predicti~n mode, the fr~me-to-frame mot;on vectors~ that is a motion vector FMVB between the I-rramc and the B-~rame and a motion vector ~VB between th~ P-frame and the B-frame, are detected, for the ~dd-numbersd cy¢le, using the t~mporally forward and backward pictures as the reference pictures, that one of the forward prediction, backward prediction and bidirectional prediction, which is a mean value between the two predictions, which will give $he last value of the estimated error from the current frame9 is selected~ and the di~erence from the original picture is en~oded, using the macro block, composed oP alternate odd-numbered and even numbered fields, as the prediction picture.
Cbnversely~ no data is transmitted for the even-numbered cycles of the current mode, in the same manner as above.
I~ the block-forming mode for the B ~rame is the frame processing mode, with the motion predieltion being of the field predic~ion mode, the temporally forward and backward pictures are used for the odd-n ~ ered eycles as the reference pic~ures, and prediction is made of the respective odd-numbered ~ields and the even-nu~bered ~ields. The respective motion vectors, namely a motion vector FMVoBo between the Io ~ield and the Bo field, a motion vector FMVeBo between the Ie fie]ld and the Bo field, a motion vector F~VoBe between the Io field and the Be field, a motion vector FMVe~e between the Ie field and the Be field, a motion vector ~VoBo between the Po field and the Bo field, a motion vec~or ~ eBo between the Pe ~ield and the Bo ~ield, a motion ve~tor ~VoBe between the Po field and the Be ~ield and a motion vector BMVeBe between the Pe field and the Re field are detected. The one of the prediction ~or the odd-numbered field, the prediction for the even-numbered field, and both predictions, for e~ample, a mean value of the two predictions, by the respeative vectors, which will give the least value of the predicted error from the current frame, is selected, ~nd the difPerence from the original picturc is encoded, using the macro-block, composed oP the Io and Ie fields or Po and Pe fields, as 2 0 J V ~_ 7 ~
the prediotion picture. Conversely, no data is transmit~ed for the even-numbered cycles of the current mode.
If the block-forming mode for the B ~rame is the field processing mode, with the motion prediction being o~ the frame prediction ~ode, ~rame-to-~rame ~otion vectors, namely a motion vector FMVB between the I-~rame and thc B-~rame and a motion vector BMVB between the P-frame and thc B-~rame, are detected~
for the odd-numb~red cycles, using the temporally forward and backward pictures as reference pictures. The ons of the forward prediction, backward prediction and bidirectional prediction, which is a mean value between the two predictions, which will give the least value of the predicted error from the current frame, is sele~ted, and the difference from the original picture is encoded, using the above-mentionod macro-block, composed of odd-numbered and even-numbered fields, separated from one another, as a prediction picture. Conversely, no data is transmitted ~or the even-numbered cycles of the current mcde.
I the block-forming mode for the B frame is the framc processing mode, with the motion prediction being of the field prediction mode, the temporally forward and backward pictures are used for the odd-numbered cycles as the reference pictures, and predic~ion is made o~ the respective odd-num~ered fields and the even-numbered fields. The respective ~)tion vectors, namely a motion vector FMVoBo be~ween the To fie]ld and the Bo field, a motion vector FMVeBo between the Ie field and ~he Bo field, a motion vector FMVoBe between the lo field and the Be field, a motion vector FMVeBe be~ween the ~e field and the Be ~ield, a motion vector BMVoBo between the Po field and the Bo field, a motion vec~or BMYeBo between the Pe ~ield and the Bo field, a n~tion vector BMVoBe between the Po ~ield and the Be field and a motion vector ~VeBe between the Pe field and the Be field are detected. The one of the prediction for the odd-numbered field, the prediction for the even-nwnbered field, and both predictions, ~or e ~ nple, a mean value Or the two predictions, by the respective vectors, which will give the least value of the predicted error rrom the current frame, is sel0cted, and the di~erence rr~n the original picture is encoded, using the macro-block, composed of the Io and Te rields or Po and Pe ~ields, as the prediction picture. Conversely, no data is transmitted ~or the even-numbered cycles of the current mode.

9~,1 7~
~ lowever, with the embodiment 1, motion predietion between Io and Ie fields, motion prediction between Po and Pe ~ields or motion prediction between lo and le fields or motion prediction between Bo and Be fields cannot be achicved, as may be seen from Fig.8.
In such case, prediction from an odd-numbered field to an even~numbered ~ield may be made in each picture by employing the embodiment 2. I~ the block-~orming is Or the ~rame processing mode, each of unit blocks Or a macro-block MB, namely luminance blocks Y0, Yl, Y2 and Y3, each composed o~ odd-numbered and even-numbered fields, and ~olor d~fferen~e blocks CbO, Crl, each composed of odd-numbered fields, is processed wi~h ~CT for the odd-numbered blocks, as shown for ex~n~le in Fig.9. If the block forming mode is the field processing mode, each of the unit blocks, that is the luminance blocks Y020, Yl30 of the odd-numbered field and the color difference blocks CbO, Crl of the ~dd-numbered field, is processed with DCT ror the odd-n~nbered cycles Subsequently, the luminance blocks Y02e and Y13e of the odd-numbered fields are processed with DCT during the even-numbered cycles.
With the present example, shown in Fig.9, motion prediction SMCI between the Io and Ie ~ields and mo~ion predi~tion SMCP
between the Po and Pe ~ields may be made in addition to the motion predictions MVP, MKoPo, MKoPe, ~ o and MCePe sh~wn in Fig.9.
Thus, in the embcdiment 2, i~ the block forming mode -~or the rame is the frame processing mode, the Io and le ~ields are cambined ~or the odd numbered cycles to form the macro-block and, ~or the odd-nwnbered cycles3 for example, DCT, quantization and variable length encoding are per~ormed ~rom one macro-block to another. It is noted that DCT is per~ormed on the above-mentioned 8 * 8 unit block basis. Conversely, no data is transmitted during the even-numbered cycles Por the current mode.
If the block-forming mode is the field processing mode, only odd-numbered Pields of the macro-block are encoded in a similar manner for the odd-numbered cycles. In this manner, at a time point of completion of the odd-numbered cycle, the Io field in its entirety and a macro-block fraction of the Ie field by the frame processing mode are obtaincd at the decoder as later explained. For the even-numbered cycle of the I-frame, motion - 2 ~ -, .. ~, . ' ' .. . . .

- 2 ~
prediction is performed of the macro-block of the le field by the field processing mOde3 using the lo ricld as a referenc~ picture, and the motion vector SMVI as well as the diffelence ~rom the prediction picture is encoded~
I~ the current Prame is the P-~ramc, the rollowing operation is performed. If, for example, the b]ock-forming mode for the P~frame is the frame processing mode and motion prediction is of ~he frame prediction mode, the framc-to-~rame motion vector MVP
is detected, using the forwardpicture (I-frame picture) as a reference picture, for the odd-numbered cycles, and a di-fference ~rom the orlginal picture is encoded, using the macro-block composed o~ Io and Ie fields as a pr~diction picture. No data is transmitt~d for the even-numbered cycles for the current mode, in the same manner as above.
~ the block-~orming mode is of the field processing mode, with the motion prediction being of the field prediction mode, a motion vector MVoPo between the lo field and the Po field and a mntion vector MVePo between the Ie field and the Po ~ield are detected, for the odd-numbered cycles, using the Io and Ie fields or the Po and Pe fislds as reference pictures. The one of the prediction for the odd-numbered fields, prediction for the even-numbered fields and b~th predictions, for example, a mean value of the prediction for the even-numbered fields and that for the odd-numbered fields, is selected9 and the difference thereof from the prediction picture is encoded. For the even-numkered cycles of the current mode, the motion vector MVoPe between the Io ~nd Pe fields, the motion vector ~VePe between the Ie and Pe ~ields and the motion vector SMYP between the ~o and Pe fields are detected ~or the macro-block o~ the field processing mode. The one o the prediction for the odd-numbered fields, the prediction for the even-numbered ~ields and the prediction for the odd-numbered fields of the current frame (motion prediction from Po field concerned only with the even-numbered cycles) by the respective vectors and the prediction which is a mean value between two of the above-mentioned predictions, which will give the least value o~ the predicted error, is selected, and the di~erence thereof from the prediction picture is encoded.
If the block-rorming mode for the B-~rame is the ~rame processing mode, and motion prediction mode is the frame prediction mode, frame-to-frame motion vectors, that is a motion vector FMVB between the I-frame and the B-frame and a motion vector ~MVB between the P-frame ~nd thc B-frame, are detected for the odd-numbered cycles, using forward and back~ard pictures as reference pictures. The one of the forward prediction, backward prediction and bidirectional prediction9 that is a mean value between forward and backward predictions, which will give the least value of the predicted error from the eurrent frame, is selected, and the difference of the selected prediction from the current picture is encoded. Converse]y, no data is transmitted for the even-numbered cycles for the current mode.
If the block-forming mode is the f;eld processin~ mode, with the motion prediction being of the field prediction mode, prediction of odd-numbered and even-nu~bered fields of the forward and backward pictures as the re~erence pictures is per~ormed for ~he odd-numbered cycles, and respective motion vectors, namely. a motion vector FMYoBo between the Io and Bo fields, a motion vector FMVeBo between the Ie nd Bo fields, a motion vector BMVoBo between the Po and Bo fields and a motion vector BMVeBo between the Pe and 80 fields are detected. The prediction which will give the least value of the predicted error is selected in the same manner as before and the difference o~
the selected pr~diction and the predicted picture is encoded.
For the even-numbered cycles of the current mode, predictions by the motion vector FMVoBe between the Io and Be fields, motion vector FMVeBe between the Ie and Be fields, motion vector BMVoBe between the Po and Be fields and ~he moltion vector BMVeBe beltween the Pe and Be fields and prediction of the odd-numbered ~ields of the current frame, that is the prediction by the motion vector SMVB between the Bo and Be fields, are additionally performed, and such a prediction which will give the least prediction error is select~d. A diference is ~hen taken of the selec~ed prediction from the predicted picture.
When handling the so-called 4:2:2 c~nponent digital VTR
format with the above~deseribed ~mbodiment 1, the unit blocks of the macro-block, ~hat is the luminanee blocks YO, Y1, Y2 and Y3, each eomposed of odd-numbered ~nd even-numbered fields, ~nd the color di~ference blocks CbO1, CrO1, Cb23 and Cr23, each composed o~ odd-nwnbered and even-numbered blocks, are processed with DCT.
For the field processing mode, the lwninance bloeks YO20 and Y130 Oe the odd-numbered ields, eolor difference blocks CbO1230 and 7 ~'3 CrO1230 o the ~dd-numbered fields~ the luminance blocks Y02e and Y13e of the even-numbered fields and color difference blocks ChO1~3e and CrO123e of the even-numbered fields, are processed with DCT.
Motion prediction ~or the example Or Fig.11 is perPor~ed in a manner as shown in Fig.8. However, with the present example, shown in Fig.11, motion prediction bctween Io and Ie fields, motion prediction hetween Po and Pe ~iclds or motion prediction between Bo and Be ~ields cannot be achieved, in the same m~nner as akove.
In such case, the embcdiment 2 may be employed, as explained previously. If the block-forming is o~ the frame processing mode, the unit blocks nf a macro-block, namely the lumin~nce blocks YO, Y1, Y2 and Y3 and color difference blocks CbO1, CrO1, Cb23 and Cr23, each composed of ofld-n ~ ered and even-numbered fields, are processed with DCT for the odd-numbered b]ocks, as shown for example in Fig.12. I~ the block forming m~de is the field processing mode, respective unit blocks, that is the luminance blocks YO20, Y130 co~posed of the od-numbered fields and the color difference blocks CkO123o and CrO1230 composed of the odd-numbered fields are processed with DC~ for the ~dd-numbered cycles. Subsequently, the luminance blocks Y02e and Y1~e o~ ~he even-numbered ~ields and the color difference blocks ChO12%e and CrO123e c~mposed of the even-numbered fields are processe~ with DCT for the even-number cycles.
Motion prediction for the e~ample of Fig.12 is the same as that shown in Fig.12.
For handling the abave-mentioned 4:2:2 component digital VTR
format by the above-described embodiments 1 and 2, an arrangement may be so made that, besides the operation shown in Figs.11 and 12, frame movement predi¢tion is performed on the basis of the macro-block MB, as shown in Fig.i3, whPreas, for field motion prediction~ a ¢ertain macro-block MB(i, j) and another macro-block MB(i+1, j) therebelow are grouped as a macro-block set MBg and motion prediction for the odd-numbered fields and the motion prediction for the even-numbered fields are per~ormed of the macro-block set MBg.
Fig.14 shows several macro-blocks o~ the fr~me shown partially in Fig.13. It is assumed that the operation proceeds in a direction shown by arrow in Fig.14. That is, Fig.14 shows ,~

2 ~

a macro-block MB(i, j~1) next to the macro-block MB(i, j), and macro-blocks. M$(i~ and MB(i~ 1) therebelow, that is, belonging to the next line.
In the macro-block shown in Fig.lA, the luminance blocks YO, Y1 and the color di~ference blocks CbO1, CrO1 are processed with DCr, with e.g. the frame processing msdc, for each of the macro-blocks MB(i, j), MB(i, j~1), ..., MB(i+1, j) and MB(i+1, j+1).
Cons~quently, with the framc processing mode, processing of each macro-block is not affected by the processing of other m~cro-blocks.
Conversely, with the ~ield processing mode, as shown in F;g.15, the macro-blocks making up the macro-block set ~Bg are divided into macro-blocks MBgo of odd-numbered fields and macro-blocks MRge of even-numbered fields and luminance blocks YOo~ Ylo and color difference blocks CbO10, CrO10 within the odd-~ield macro-block MRgo are processed with DCT. If the macro-block set MBg is made up of the macro-blocks MB(i, j~ and MB~i~1, j) of Fig.14, the luminance blocks YOo, Ylo wi~hin the odd-field macro-block MBgo in the macro-block MBg are made up of odd-field luminance blocks of the macro-block MB(i, j) and odd-field luminance bloc~s of the macro-block MB(i~1, j3 and the color difference blocks CbOlo~ CrOlo within the odd-field macro-block MBgo are made up o~ odd-~ield cclor differen~e blocks of the macro bloc~ MB(i, j) and odd-~ield ~olor difference blocks o~ the macro-block MB(i~13 j). Similarly, the luminance blocks YOo, Ylo within the even Pield macro-block MBge in ~he macro-block MBg are made up o~ even-field luminance blocks oP the macro-block MB(i, j) and even-field luminance blocks of the macro-block MB(i~1, j) and the color dir~erence blocks CbOle, CrOle within the even-field macro-~lock MRge are made up of even-~ield color difference blocks of the macro-block MB(i, j) and even-~ield color dif~erence blocks of the macro-block MB(i+1, j).
As amy be seen from above, ~he relation between motion prediction and DCT processing modes is as follows: In the -present encoding device, if, in the macro-block MB(i, j), the motion pr~diction is of the frame processing mode and the DCT is o~ the frame processing mode, the dif~erence between the predicted picture and an input picture (original picture), extracted ~rom a picture decoded in the ~rame memory group fitted with motion compensator 20 as a reference frame, ~s processed . , , ~ .

2 ~ 9 ~ 1 ~3 with DCT. The DC~ coefficien~s and frame motion vectors are transmitted.
If~ in the macro-block MB(i, j), the motion prediction is of the field processing mode and the DCT ;s of the field processing mode, the difference between the predic~ed picture extracted ~rom the odd-numbered field and the odd-field original picture and the odd-~ield motion vector are encoded for the macro-block MB(i, j), while the dirrerence between the predicted picture extracted ~rom the even-numbcr~d ~ield and the even~field original picture and the even-field motion v~ctor are encoded for the macro-block MB(i+l, j).
If, in the macro-block MB(i, j), ~he motion prediction is o~ the field processing mode and the DCT is of the ~rame processing mode, a frame difference betwee~ a predicted picture for the position of the macro-block MB(i, j) taken out from the re~erence ~rame ~nd the input picture, the odd-field motion vector and the even-field motion vector are transmitted for the macro-block MB(i, j). A frame dif~erence between a predicted picture ~or the position o~ the macro-block MB(i+l, j) taken out from the reference ~rame and the input picture is transmitted for the maoro-block MB(i+l, j).
If9 in the macro-block MB(i, j), the motion prediction is of the frame pr~cessing mode and the DCT is of the field processing m~de, a difference between a predicted picture taken out Prom the odd-numbered field and the odd-field original picture, the ~rame motion vector o the macro-block MB(i, j) and the ~rame motion vector of the macrorblock MB(i+l, j) are transmitt~d for the ma~ro-block MB(i, j~, while the differenoe between the odd-numbered predicted picture and the input picture is transmitted for the macro-block MB(i~l, j~.
Meanwhile, with the enc~ding device of the present embodiment, the present ccde is realized by adding extension bits to the oonventional macro-block type for interchangeability w;th the cnnventional system.
That is, in the embodiment 1, the m~cro-block type has three predictions, n~nely forward prediction, backward prediction and bidirectional predictiun. Since prediction from odd-numbered and even-numbered ~ields of the previous frame are possible with the forward prediction, the present code is realized by appending extension bits useful for recognizing which of these predictions -- 2 ~ --is used. Since there are two predictions, i~ suffl~ces to add one bit for one dir~ction, that is for thc forward or backward direction. For example, if the prediction is forward or backward prediction and from the odd-numbcr~d fields, it s~ffices to add a code 1 as an extension bit to the conventional macro-bit type, Similarly, if the prediction is ~rom the even-numbered field, it suffices to add a code O as an extension bit to the conventional macro-bit type. For bid;rectional prediction, bo~h extension bits are appended ~or b~th forward and backward predictions.
Meanwhile, ~or a ~rame prediction mode, no extension bits are appended, so that the bit string is of the same style as the conventional bit stream (MPEG).
The above applies to ~he P-~rame as well.
In the embodiment 2, the macro-block type has three predictions, namely forward prediction~ backward prediction and bidirectional prediction for e.g.the B-~rame. If the prediction is the forward prediction with the field prediction mode, it is necessàry to append extension bits to the macro-block for allowing to recognize if, when the prediction is for the field prediction mode, the prediction is that from an odd-numbered field, an e~en-numbered field or from an odd-numbered field of the current frame. That is, since there is a prediction from the current frame in the forward field pred]lction mode, one or two extension bits are required for express;ng three predictions including odd- and even-numbered predictions by extension bits.
On the other hand, for backward field prediction mode, since there are two predictions, that is predic~ion or odd and even numbers, 1 extension bit is required. For example, for forward prediction, a code 1 for odd-odd-nu~bered field of a forward frame and, a code Ol for even-numbered field of a forward fr~me a~d a c~de 11 ~or prediction from odd-numbered field of a rear frame are appended. For backward prediction, it su~ices to append ~odes 1 and O for prediction from an odd-numkered field of a backward frame and for prediction from an even-numbered field of a backward frame to the conventional macro-bit type.
If the mod~ is the frame prediction mode, no extension bits are appended, so that the bit string is o~ the same style as the conventional bit stream (MPEG). For bidirectional prediction, both extension bits are append~d for forward or backward prediction.

.

2 ~ 7 ~
The above applies to the ~-frame as well.
As a mcdification, the number of the extension bits can ~e reduced to one for forward prediction Tha$ is, in the even-numbered cycle for the field prediction mode, the number of bits may be reduced to two and the forward prediction mode may be transmitted by one extension bit by abolishing prediction from the odd-numbered field furthest remov~d both temporally and spatially, as shown in Fig.16. Specifically, or odd-numbered cycles and for forward prediction, it su~ices to add codes 1 and O to the conventional macro-block type as extension bits in the ~ase of prediction ~rom an odd-num~ered ~ield o~ the forward frame and an even-numbered field of the ~orward ~rame. Similarly, for even-numbered cycles and for forward prediction, it sufices to add codes 1 and O ~o the conventional macro-block ~ype as extension bits in the case of prediction from an odd-numbered field of the current frame and an even-numbered field of the forward frame. Finally, for backward prediction, it suffices to add codes 1 and O to the conventional macro-block type as extension bits in the case o~ prediction from an odd-numbered field of the backward frame and an even-numbercd field of the backward frame.
The block diagram of Fig.17 shows a decoder (first and second decoding units) for picture signals as a counterpart device of the encoding device oP the akove described embodiments 1 and 2. The high efficiency dec~der includes a variable length encoding circuit 51 ~or receiving and dccoding coded data o a picture to be reprodu~ed and the header in~ormation for outputting the detected motion vector information, the motion prediction mode information indicating which of the frame-by-frame motion eampensation and the field-by-~ield motion e~mpensation in a maero-block is more effieien~, the block-forming mode information indicating which o~ the block ~ormation ~or frame-by-~rame orthogonal transform and block ~ormation for frame-by-fr~ne orthogonal transform in a maero-block is more effieient, and a macro-block address increment in the maero-block header information. The high effieieney decoder also ;neludes address generators 81, 82 and 83 by caleulating address increment values at ~rame buffers 61, 62 and 64 storing pieture deeoding data fr~n the maero-address inerement for finding leading addresses of the maero-blocks and supplying the l~ading addresses 2 ~
to the frame buffers 61, 62 and 64, ~nd motion compcnsation circuits 59, 60, 63, 65 and 66 for supplying the relatiYe addresses of the macro-blocks other than the leading addresses to the frame buffers 61, 62 and 64 for data accessing, receiving the detected motion vectors, motion pr~d;ction mode information and the block-forming mode information, executing the compcnsation between motion-compensat~d frames or fie]ds in association with the mode information and transmitting the motion-compensated picture information to the frame bu~fers 61, 62 and 64.
Referring to Fig.17, data encoded by the high efficiency enc~ding devioe of the embodiments 1 and 2 are transiently recorded on a storage medium, such as CD. The enccded data reproduced form CD is supplied via input terminal 50 to an inverse variable length encoding circuit 51 where the header information etc. are decoded ~rom one s~quence to another, ~rom one frane group to another and from one frame to another. The header information is deccded from one slice (macro-block gro~p) to another. The quantization ~idth is included in the slice header. The macro-block type indicating the macro-bloc~ ~ ~
addresses, frame processing mode/field Iprocessing mode and the decoding system is decoded from one macro-block to another. The quantiza~ion width is de ~ ~d at ~he ~ime of the updatin~.
Meanwhil~, if the block formation in a macro-block is o the frame processing mode, the macro-block in its entirety is decoded for the odd-numbered cy~le, while no data is decoded for an even-num~ered cycle. If the block formation in a macro-block is o~
the ~ield processing mode, only the block including odd-numbered ields in the macro-block is decoded for the odd-numbered cycle, wile the block including even-number~d fields is decoded for the even-numbered cycles.
The picture information is decoded by a dequantizer 53 performing an op~ration which is the reverse of quantization and an inverse DCT circuit 54 performing an inverse DCT operation, and a decision is given by the macro-block type as to whether or not the information is a differential information. Depending on the results of decision, a mode switch 57 for changeover between addition or non-addition to the re~erence picture (or between non-intra/intra of MPFlG ccding) by an additive node 56. The decoded picture is entered to frame buf~ers 64 or 61 for the I-or --^` 2 0 ~
P-frame (alternately each time the l-frame or the P-frame s processed) ~nd to frame buffer 62 i~ ~hc current frame is the B-frame. Each frame buffer is composed o~ two fie]d buffers and the odd/even Pield p;cture is stored separately in each o~ these field buffers. Recording in the framc bu~fers is controlled by changing over a switch 58.
The addresses r~corded in the frame buffer are supplied from an address generator. The address generator calculates an address increment value in the frame buffer from the macro-block address increment in the header information of the macro-block to find the leading address of each macro-block.
l~e quantization width data is stored in a one-field memory 52. The quantization width data is supplied to dequantizer 53 via switch 55 changed over depending on the output of the inverse variable length enccding circuit 51. Since only the macro-block processed with the field processing mode is decoded for the even-numbered cycles, macro-block addresses decoded for each ~acro-block9 macro-block type and the motion v~ctor required by the prediction system indicated thereby are decoded and the differential picture from the reference field is added to the motion-compensated picture to produce the playback picture.
In the case of an encoding system in which the even-field quantization width data are transmitted independently o~ the odd-numbered field for the even-numbered cycle, it is unnecessary to store the quantization width data transmitted for the odd-numbered cycle, so that the one-~ield ~mory 52 may be eliminated.
On the other hand, ~ata o~ the frame buffers 64, 62 and 61 is motion-compensated by the motion compensation circuits 65, 66, 59, 60 and 63. At this time, the respective motion compensation circuits change over the frame motion compensation/field motion compensation (frame/field) depending on the block-~orming mod~
at the time oP thP DCT operation.
These motion-compensated pictures are transmitted to respective fixed terminals of changeover switches 67, 68 and 71.
These changeover switches 67, 68, 71 are changed over ~or taking out the re~erence field or frame indicated by a de~cding system o~ the macro-block decoding ~ype. The changeover switch 71 is supplied with an output of switch 67 and a signal which is outputs o~ the changeover switches 67 and 68 added together by ~?J~QJ 7 ~
additive node 69 and subsequently halvcd by a divider 70. An output of switc~ 71 is supplie~ to switch 57.
Outputs of frame buffers 64, 61 ad 62 are supplied via changeover switch 72 to a display 73. The display 71 is supplied with the outputs of the changeover switeh 72 which are changed over in the sequence of playback pictur~s instead of in the decoding sequence for producing an output picture.
It will be seen from a~ove that~ i~ there is a moving object CA ahead of the stationary background, as shown in Fig.46, the moving object is seen as being zig-zag shaped, as shown at KS, i~ one frame is viewed, because ~here is a movement between fields. With the device of the present emhodiment, since the moving object is encoded with the field processing mode, it can be processed as a picture devoid o~ deviations so that a high-quality moving picture may be reproduced by odd/ev~n motion compensation. Thus, as shown ~or example in Fig.18, the moving portion is processed with the field processing mode during odd-numbered cycles, whereas the stationary portion is processed wi~h the frame processing mode, as shown for example in Fig.18.
Meanwhile, if a picture is already formed during the even-numbered cycle9 the stationary portion is as shown by hatched lines in Fig.19. In Fig.19, the portion other than the hatched porion, that is the moving portion9 is decoded by motion compensation.
Meanwhile3 with the res~nt embcdimenlt, sinee only the macro-block processed with the field processillg mode is decoded during the even cycle, i~ is necessary to know a macro-bloek address.
There are two methods ~or grasping the rnacro-block address. One is to transmit the address of the macro-block for each macro-block of the even cycle as described previously. The other method is to store the information of the one-ield processing mode/rame processing mode information during the odd-numbered cycle and to calculate the address o~ the macro-block in the field processing mcde rom the string of each processing mode.
rrhe former method has an advantage that it is necessary to add memories, whereas the latter method has an advan~age that the transmitt~d information is not increased in volume. The same may be said of the quantization width if the quantization width is transmitted for each macro-block without storing one-~ield data during the above-mentioned odd~numbcred cycle.

~v~.7~
With the above-described embodimcnts 1 and 2, since one-frame processing is divided into two cycles, namely an odd cycle and an even cycle, the frame processing mode is changed w er to the field processing mode or vice versa on the macro-bl~ck basis during the od-numbered cycle, both thc odd-number~d field and the even-numbered cycle are decoded during ~rame processing~ only odd ~ields are decoded for field processing, the quantization width for the cycle is stored, and the storcd information is used during the next even cycle for motion-compensating only the macro-block of the ~;eld processing mode ~or decoding the playback picture, encoded data may be transmitted highly efficiently. That is, the high-quality moving picture may be reproduced with a smaller volume of the transmitted information.
By way of embodiments 3 to 6, a high efficiency encoding device for picture signals ac~rding to the present invention and third to sixth decoding devices associated therewith will be explained in detail.

*** THIRD EMBODIMENT ***
The block diagram of Fig.20 shows a high efficiency encoding device according to the embodi~ent 3. In ~his figure, the blocks indica~ed ~y the same n~merals as those used in Figs.1 and 5 are operated in a similar manner. Therefore9 only the blocks bearing different numerals ~rom those used in Figs.1 and 5 are explained. The high effic;ency encoding device of Fig.20 includes, besides the blocks bearing the same numerals as those shown in Figs.1 and 59 an encading mode decision circuit 34(a) and a selector ~4, as limitation mode selecting means, for selecting a irst limitation mode of inhibiting encoding by the above-mentioned frame processing mode for all of the macro-blocks in each frame or a second limitation mode o inhibiting prediction of an even field of a current frame being encGded from an odd fiel~ oP the same frame for the entire macro-blocks in one frame, whichever is more eff;cient, and an address generator 35(a) for controlling a frame memory group to output odd-numbered field ~omponents oP the entire macro-blocks if the first limitation mode is elected for one frame or picture ~nd subsequently to output even-numbered field components of the entire macro-blocks, as well as to sequ~ntially output the m2cro-blocks, on the frame-by-frame basis, if the second limitation mode is selected, based on a frame constituted by odd-numbered and even-numbered field components of the macro-blocks as a unit.
That is, the high efficiency encoding device of the embodiment 3 includes, for encoding a moving picture hav;ng a frame made up Or two ~rames~ encoding m~ans for dividing each of the blocks of the frame into an odd field (first field) and an even field (second ~ield) and rendering the motion prediction of the first tn second fields possible (first limitation mode~ and for changing over between first and second field division/~irst and second field non-division on the macro block basis by way of block formation (second limitation mode). These encoding mcans are changed over ~rom frame to frame. A l-bit information indicating these encoding means, that is the information indicating the selected msde, is append~d to the codes.

2 ~ 9 9 ~ 7 ~

*** FOURTH EMBODIMENT ***
The block di~gram o Fig.21 shows a high efficiency encodin~
device a~cording to the embodiment ~. Tn this figure, the blocks indicated by the same numerals as those used in Figs.l and 5 are operated in a simi]ar manner. Therefore, only the blocks bearing different numerals from those used in Figs.l and 5 are explained.
The high e~ficiency encoding device of Fig.20 includes, besides the blocks bearing the same numera]s as those shown in Figs.l an~
5, an encoding m~de decision circuit 3~(b) and a selector 24, as limitation mode selecting means, for selecting a first limitation mode of inhibiting encoding by the above-mentioned frame processing mode for all of the maoro-blocks in each slice or a second limitation mcde of inhibiting prediction an even ield o a frame being encoded from an odd field of the same frame for the entire macro-blocks in one slice, whichever is more efficient, and an address generator 35(b) ~or controlling a ~rame mcmory group to out~ut odd-numbered field components o the entire macro-blocks if the first limitation mode is selected for one frame or picture and to output even-numbered field components for the ~ntire macro-blocks, as well as to sequen~ially outpu~, if the second limitation m~de is selected, the macro-bloc~s, by one slice at a time, based on a framc constituted by odd-numbered and even-numbered field components of all of` the macro-blocks as a unit.
That is, the high ef~iciency encoding device of the embodiment 4 includes, or enc~ding a m~vlng picture having a frame made up o~ two f~ames, encod;ng means for dividing each of the blocks in the frame into an odd Pield (~irst field3 and an even field (second field) for rendering the motion prediction of the first to second fields possible (Pirst limitation mode) for changing over between first and second field division/first and second ~ield non-division on the macro-block basis by way o~
block formation (second limitation mcde). These encoding means are changed over from ~rame to frame. A l-bit information indicating these en~oding means, that is the information indicating the selected mode, is appended to ths codes.
Referring to the drawings, the embodimer.ts 3 and 4 are explained in detail.
Fig.20 shows a third high efficiency encoding device for ~ ~ 9 J~ 1 r~) 3' picture signals according to the e~bodimcnt 3 of the present invention. With the present encoding device~ encoding is performed on the basis of macro-b~ocks each consisting in a two-dimensional array of plural pixels sm~ller in size than a picture. For example, each macro-block consists of 16 16 pixels in a spatial array o~ input pic~ure data in a ras~er scanning sequence.
~ rhe high ef~iciency encoding device of the embodiment 3 ineludes a frame membry group 10 for storing~ as an original picture, a ~rame (picture) consisting of plural unit blocks (macro-blocks) each consisting of 16 * 16 pixels, and motion detection means made up o~ a frame motion detection circuit 22 as means ~or detecting the sum o~ absolute values of the pixels and motion vectors between the frames on the macro~block basis.
~nd a ~ield mo~ion detection circuit 21 for detecting on the macro-block basis, the sum of absolute val~es of the pixels and motion vectors between the fields, made up o~ even-numbered and odd~numbered s~anning lines of the frame pixels.
The device o~ the present embodiment also includes a frame/field mode decision circuit 33 made up of first mode selccting means and sec~nd msde selecting means. The first mode selecting means decides which of a frame prediction mode of carrying out motion compensation bascd on a frame in the macro-block or a field prediction mcde of carrying out motion compensation based on a ~ield in the macro-block is more efficient and selects the more efficient mode. The second mode selecting means d~cides which of a frame processing mode of forming blocks for carrying out orthogo~al transform, such as DCT, based on frame in the macro-block, of a field processing mude o~ forming blocks for carrying out orthogonal trans~orm, such as DCr, based on of a field in the macro-block is more efficient ~or carrying out orthogonal transform, using the output information o~ the motion detectîon means and the first mode selecting means.
Besides, ~he present embodiment 3 includes, in addition to the motion detection means and the frame/field n~de decision circuit 33, a limitation mode decision circuit 34(a), as limitation mode selecting means, for deciding which of the second limitation mode o~ adaptively changing over the mode o~ block formation for orthogonal transrorm between the frame processing "

- 2 0 ~ v~ ~7,3 mode and the field processing mode on the basis of each macro-bloc~ in each framc for encoding ~ach macro-block in accordance with the selected mode or the ~irst limitation m~de o~ ~orming thc blocks for orthogonal transform of the entire macro-blocks in each frame in accordance with the fic]d processing mode, encoding odd fields in the macro-b]ocks in an amount corresponding to one frame during odd-numbered cycles and encoding even fields in the macro-blocks in an amount corresponding to one frame during cven-numb~red cycles is more cfficient for encc~ling and selecting the more eeficient limitation mode. The odd-numbered and even-numbered cycles m the periods of scanning of odd-numbcr~d and even-numbered fields in interlaced scanning, respectively.
:. Meanwhile, the decision circuit 34(b) of the fourth encoding device shown in Fig.21 includes limitation mode selecting means for deciding which of the s~cond limitation mode of adaptively changing over the mode of block formation for orthogonal transform between the rame pr~cessing mode and the field pr~cessing mode on the basis of each macro-block in each frame for encoding each macro-block in accordance with the selected mode or the first limitation mode of forming the blocks for orth~gonal tr~ sform of the entire macro-blocks in each slice in accordance with the fiel~ processing m~de, encoding odd fields in the macro-blocks in an amount corresponding`to one frame during odd-numbered cycles and encoding even fields in th~ macro-blocks in an amount correspondi~g to one frame during even-numbered cycles is more e~ficient ~or ~lcoding and selecting the more eficient limitation mode. The odd-numbered ~d even-nwnbered cycles mean the periods of sc~ulning of odd-numbered and even-numbered fields in interlaced scannin~, resp~ctively.
Fig.42 shows a modification of limi~ation mode selecting means o~ the embodiment 3. Wi~h the third encoding device, data FDAD and FMAD, as ~ound from macro-block to macro-block, are cumulated ~rom frame to rame to find cumulated data SFDAD and SFMAD. hhen the cumulative data SFDAD becomes smaller than FMAD
+ T, T being an offset value, the second limitation mode is selected, and otherwise, the first limitation mode is selected.
Besides, with the third encoding device; data FDAD and FMAD, as found fr~m macro-block to macro-block, are cumulated from sli~e to slice to find oumulated data SFDAD and SFMAD. When thè

i - 3 9 -2 ~ 3 . . , cumulative data SFDAD becomes smaller than FMAD ~ T, T being an offset value, the second limitation modc is selected, and otherwise, the first limitation m~de is selected.
Similarly, a flow chart Or Fig.23 shows a modification of mode decision in the limitation means sclecting means of the emb~diment 3. With the third encoding device, the limitation mode sele~tion is made using a motion vector from the odd field (first field) to the even field (second field) oP the current ~rame being encoded. In Fig.24, the motion vector MV ~rom this odd field to the even field, shown in Fig.24 by ~otion vector Mv1-2.
In the flow chart of Fig.23, the motion vectors are found at step S21 for all o macro-blocks in the current ~rame. At step S22, a median value between a horizontal component (x)and vertical component (y) of each motion vector is found in the following manner. First, the horizontal components of the motion vectors are arrayed in the order of falling power. The value of the mid data becomes the median value Mv_ x. The median Mv_ y o~ the vertical components is fo~nd in a similar manner.
The vectors MV (Mv_ x, Mv_ y), thus found, represent parameters showing the motion of the picture in its entirety.
The magnitude r o~ the vector MY is inl~roduced as a parameter indicating the magnitude of he motion of the entire picture. The magnitude r may be found by the equa~ion (3).
~ Equat 1 on 3 ~
r = ¦MV¦ = ~qrt ~MV_X^2 + MV_Y^2) ~3) ; At step S24, the limitation mode is changed over depending on the magnitude r. Since the first and second limitation modes are more meritorious for pictures having faster and slower movements, respectively, the second and rirst limitation modes are selected i~ r is not more than a certain threshold and otherwise, respectively.
That is, the second limitation mode and the first limitation mode are selected i~ r < threshold and r > thresho]d, respectively.
- Meanwhile, with the fourth encoding device, the limitation mode selection is made using a motion vector from the odd field (~irst ~ield) to the even field (second field) of the current - frame being encoded. The motion vectors Mv are found from the .. . .. . .
, ' ,: ' ~ . ' " : ' 2 ~
even fields to the odd fields for all o~ the macro-blocks in he slice being encoded, and median values MV (Mv_ x, Mv_ y) of the horizontal ~nd verti~al c~mponents thereo~ are found. Similarly~
the above magnitude r is found, and th~ second limitation mode is selected if the value r is not more than a certain threshold value. If otherwise~ the first limitation mode is selected.
The flow chart o~ Fig.25 is a modi~ication o~ mode selection by the limitation mode selecting means of the embodiment 3. The limita~ion mode is selected using the correlation between the odd and even fields of the current frame bcing encoded.
The correlation between the odd and even fields is achieved by the method shown in Fig.25. This is the meth~d well-known as a method for selecting the macro-block mode in the international standardization Or the moving picture encoding and compression now under way at ISO/IEC JTCl/SC2/WGll. With the present embodiment, this method is extended and used for selecting the fr~me selecting mode.
In the flow chart shown in Fi~.25, var 1 and var 2 are first found at step Sl. Then, at step S2, the number o~ macro-blocks in the current ~rame satisfying the relation var 1 >= var 2 +
offset is found. This number is termed num_ Pield_ mb.
For the macro-block satisfying ~he rela~ion var 1 >= var 2 + offset, which h2s higher correlation between fields, the first limitation mNde is preferably employed. Therefore, the second limitation mode is selected a~ step $3 if num_ field_ mb is not more than a certain threshold, and the f~irst limitation mode is elected otherwise, for further processing.
That is, i~ num_ eield_ mb <~ threshold, the limitation mode is the second limitation mode of step S5 and, if num_ ~ield_ mb >
threshold, the limitation m~de is the first limitation m~de o~
step S4.
Meanwhile, with the fourth en~oding device, the number o~
macro-blocks num_ Pield_ mb satisPying the relation var 1 >= var 2 + of~set in the slice being encoded is found in a similar manner and ~h~e limitation mode is sel~cted depending on this value. I~ num_ Pield_ mb is not more than a certain threshold, the second limitation mode is selected and, if otherwise, the ~irst limitation mode is select~d, ~or Purther processing.
The rlow chart oP Fig.26 is a modiPication oP mode selection by the limitation mode selecting means Or the embodiment 3. At A r~?~
step S11, the difference between the previously decod~d picture referred to by the motion ve~tor and the ~urrent picture, ~or each o~ the macro~blocks o~ the current picture, is ~ound, the square sums of the differences are found, and the limitation mode selection is made at step S12, using the ~hus found square sums.
The square sums of the differences are found with the first and second limitation modes, and the limitation mode having the Iesser values Or the square sums is sel~cted.
Similarly, with the fourth encoding device, the square sums of the dif~erences in the slice being encoded are similarly ~ound, and the limitation mode having the lesser value of the squared sum is selected.
-~ The flow chart of Fig.27 is a modification of mode selection by the limitation mode selecting means o~ the embodiment 3. The limitation mode is selected using the correlation between the odd and even fields of the current ~rame. At step S51 o~ the flow chart of Fig.27, var 1 and var 2 are first ~ound.
At step S52, ~he values var 1 and var 2 are sunm~l together for all of the macro-blocks present in the current frame. At s~ep S53~ the limitation mode is selected base~ on the thus ~ound - values var 1 ~nd var 2. If var 1 >= var 2 + offset, the first limitation mode is selected and, i otherwise, the second limitation mode is selected.
In the ~ourth encoding device, var 1 and var 2 are summed -` for all of the macro-blocks in the slice being encoded to find Var 1 and Var 2. ~he limitation mcde is selected ~rom the relation between Var 1 ~nd Var 2. If Var 1 <= Var 2 + offse~, the ~irst limitation ~ode is selected and, i~ otherwise, the second limitation mode is selected.
;Similarly~ the flow chart of Fig.28 is a modifi~ation of mode selection by ~he limitation mode selecting means of the embodiment 3. With the third encoding device, limitation mode selection is made using the motion vectors and the correlation between the first and se~ond fields o~ the current ~rame. At : step S31, the motion vectors o~ each macro-block~ as ~ound at s~ep S31, are converted into unit vectors [n_ x[i], n_ y[i]. If the motion vectors are (mv~ x, mv_ y), the following equations (54) and (5) hold:
Equ~t 1 on 4 ~
n_x = mv_x / sqrt (mv_x^2 + mv_y^2 ) (4) ' /

2~v~ ~ ~J
[ Eq~J~ti on ~ 3 n_y = mv_y / ~qrt (mv_x^2 + mv_y~2) (s) At step S33, the s~m vector SMY(S_ x, S_ y), as found by summing all o the unit v~ctors, is found. At step S34, the magnitude of the sum vector ~V dividcd by the number of the macro-blocks num_ MB9 as indicated by the equation [6), is denoted as R.
t Equ~ti on ~ ~
R - ( S_x ^ 2 ~ S_y ^ 2 ) nLJm_MB
= (Sum (n_x [i~) ^ 2 + ( Sum ( n_y [; ~ ) ^ 2 ) / num_M8 (5) The value R is a statistic quantity employed in veri~ying vector anisotropy. For example, if the motion vector exhibits anisotropy, that is if the picture in its entirety is moved significantly, the value R assumes a larger value.
The limitation mode is set at stcp S36 from ~he re]ation betw~en the value R and Var 2 as ~ound from the flow chart of Fig.27. For example, if Var 2 is no$ more than a certain threshold and R is not more than a certain threshold, the second limitation mode is sele¢ted ~nd, i~ otherwise, the first limitation mode is selected (step S37).
Meanwhile, with the fourth encoding device, R and Var 2 are similarly found in the slice being enc~ded for selecting the limitation mode. If Var 2 is not more than a certain threshold and R is not more than a certain threshold9 the second limitation mode is selected and, if otherwise, the ~irst limitation mode is selected.
The device o~ the present embodim~nt includes an address generator 35 for recognizing i he cycle is an odd cycle or an even cycle and controlling the ~rame n~ory group 10 for outputting the macro-blocks divided into blocks in association with the block-~ormin~ modes ~or orthogonal transform for the odd cycles i~ the limitation mode is the s~cond limitation mode and also controlling the frame memory group 10 for outputting the macro-blocks divided into blocks in association with the block-forming modes for ortho~onal tr~nsform for the odd and even cycles i~ the limitation mode is the first limitation mode, and a ~rame memory group ~itted with a motion compensator 20, as motion compensation means, for receiving the processing mode information selected by the processing mode selecting means , "
''' ', '-' ~ "' ,'"; ; , ' , ~ ':
. ~ ' ' , "

2 ~
(frame motion prediction frame orthogonal transform / field motion prediction field processing mode data~ and for executing the motion-compensated interframe or interfield prediction responsive to the mode information.
With the present enooding device, three encodings are possible9 namely the intra-frame encoding (I-frame or I-picture), predictive interframe encoding (P-frame or P-p;cture) and bidirectional interpicture coding (B-~rame or B-picture~, as sho~l in Fig.44. E~ch picture is divided into blocks each consisting of 8 * 8 pixels, with 2 * 2 blocks, that is lB * 16 pixels, making up ~ ch macro-block.
It is noted that, with the encoding device of the embodiment 3, the above-mentioned first mcde selecting means select which of the frame predictive mode or the field predictive mode is more e~Picient for motion compensation, whereas the above-mentioned ~irst mode selecting means select wh;ch of the ~rame processing mode or the ~ield processing mode is more efficient ~or orthogonal transform. Meanwhile, the selection of the first and the second modes is perform~d by the above-mentioned frame/field processing mode decision circuit 33.
With the encoding device o~ the em~xxliment 3, not only the mode selection is made by the processing mode selecting means, but also the encading is performed for each frame in accordance with one of the two pr~cessing modes which is more efficient.
That is, with the first limitation mQde, direction into the blocks for orthogonal tr~nsform of the entire macro-blocks in each frame is carried out in accordance with the field processing m~de, only the ~dd fields in the macro-blocks are enccdcd in an amount corresponding to one frame during odd-numbered cycles and the ev~n fields in the macro-blocks are encoded in an amount corresponding to one frame during even-numbered cycles. The odd-numbered and even-number~d cycles mean the periods of scanning of odd-numbered and even-numbered fields in interlaced scanning, respectively. With the second limitation mode, each macro-block is encoded by adaptively changing over between the frame processing mode and the field processing mode for ~ach macro-block within a ~rame. The ]imitation mode selecting means decide which o~ these first and second limitation modes is more efficient for encoding and the more efficient limitation mode is selected.

- 4 ~ -2 ~ V ~ ~_ 7 ~
That is, with the above-mentioncd s~cond limitation mode, the mode of dividing each frame into blocks with~ut dividing the frame into a first field or odd ~ield ~nd a second field or even field followed by encoding (the abovc-m~ntioned frame processing mode) is adaptively changed over to the mode o~ dividing each frame into the first and second rields and dividing the fields into blocks followed by encoding (thc above~mentioned field processing mode) or vice versa in such a manner that the fr~me and field processing modes are used for a macro-bloc~ presenting small picture movements and a maero-block presenting signiPicant picture movements, respectively.
Consequently, if the frame processing mcde is selscted for the second limitation mode~ motion prediction is made from the forward and backward frames for motion prediction for the P and B frames, and the differential picture from ~he prediction-coded picture is processed with DCT. On the other hand, if the field processing mode is selected for the second limitation mode, motion prediction for the P and B frames is made ~rom the first and second fields of the forward and backward frames for each of ~he firs~ and second ~ields of the macro-blocks, and the differential picture from the prediction picture is pr~cessed with DCT. From this it may be said that the second limitation mode represents en~oding without intra-frame prediction, interframe. Besides, with the second limitation mode9 encoding is performed within the odd-numbered cycles. Meanwhile, the second limitation mode may be dePined as being intra-frame interfield encoding.
With the s~cQnd limitation mode, motion prediction between ~he fields within a frame, that is between odd and even fields within the same ~rame9 cannot be made.
In this consideration~ with the first limitation mode of the embodiment 3, division into blocks of all of the macro-blocks within each frame for orthogonal transform is made with the field processing mode. Specifically, only the odd fields in each macro-block are encoded for the odd-numbered cycles in an amount corresponding to one frame and subsequently the even fields in each macro~block are encoded for the even-numbered cycles in an amount corresponding to one frame~ Thcrefore, with the present first limitation mode, since the odd fields (first fields) are encoded first, motion precliction for the even fields (second -- ~ 5 --2 ~
~ields) may be made from the odd ~iclds (first fields~.
Meanwhile, it may be said from this that the first limitation mode is frame encoding wi~h intra~frame interfield prediction.
Returning to Fig.20, the main flow Or picture data to be encoded by the encoding device Or the ~mbodiment 3 is explained by referring to Fig.20.
In this figure, digital picture signals are supplied to input terminal 1 so as to be stored in frame memory group 10.
The ab w e-mentioned lB 16 pixel unit macro-block data are read from frame memory group 10, under control by an address generator 35 as later describ~d, and transmitted to a difference detector 12. ~he difference detector 12 is also supplied with motion-compensated picture dat~ from the frame memory group fitted with motion compensator 20 as later described and a di~ference therebetween is detected by the dif~erence de~ector 1~.
An output of difference detector 12 is supplied to a DCT
circuit 13 for orthogonal trans~orm (DCr). The DCT coefficient data, produced by DC~ by DCT circuit 13~ is supplied to quantizer 14. Quantized data ~rom quantizer 14 is outpu~ted at output terminal 2 as encoded data via a buffer 16 and a variable length encoding circuit 15 for performing variable length encoding, such as Huffman coding or run-length coding.
The ~ramR memory group ~itted with motion compensator 20 is supplied with quantized data from quantizer 14 via a dequantizer 17, an inverse DCT circuit 18 and an additive node 19. The additive n~de 19 adds the output of the inverse DCT circuit 18 to the output of the frame mcmory group ~itted with motion compensator 20. Me~nwhile, the information or inhibiting overflow of the buffer 16 is fed back to the quantiz~r 14.
On the other hand, picture data outputted on ~he macro-block ~asis ~rom the frame memory group 10 is transmitted to the ~rame motion detection circuit 22 and the field motion detection circuit 21.
The ~rame motion detection circuit 22 detects the sums of the dif~erences of absolute values of the pixels and the motion vectors between the frames on the macro-block basis and outputs the data (frame-tor frame motion vector data FMMV and data of the sums o~ the di~ferences of the absolute values FMAD). On the other hand, the field motion detection circuit 21 detects the 2 0 ?J ~J ,o_ ~ 3 sums of the di~rerences of absolute valucs of the pixels and the motion vectors between the fields on the macro-block basis and outputs t~e data ~field-~o-field motion vector data FDMV and data of the sums of the differences of the absolute values FDhD~.
The motion vector data FMMV/FDMV of the respective motion vectors of these motion detection circuits 21, 22 are transmitted to selector 2~.
The data of the sums of the difrerences of the absolute values FMAD/FDAD and motion vector data FMMV/FDMV from the frame mo~ion detection circuit 22 and the fie]d mo~ion detec~ion circuit 21 are also supplied to the frame/fiel~ mode ~ecision ci~cuit 33.
The framc/field mode decision circuit 33 decides~ at the time of motion prediction by the frame m~mory fitted with motion compensator as later explained, which o~ the frame-by-frame motion prediction or field-by-field motion prediction is to be per~ormed, based on the data of the sums of the differences of the absolute values FMAD and FDAD PrQm the frame motion detec~ion circuit 22 and th~ field motion detection circui$ 21, and outputs data indicating a prediction mode of the more efficient mode.
Specifically, i~ it is ~ound by the fr~ne/ield mode decision circuit 33 that the difference between the data FMAD and F~AD
is larger than threshold Tl ~FMAD - FDAD > Tl), the circuit 33 outputs data (data MPFD o~ the field pr~diction mode in the motion prediotion) indicating that field-by-field motion prediction is more ef~icient. Conversely, if it is found by the frame/~ield mcde decision circuit 33 that the difference between ~he da~a FMAD and ~lDAD is equal to or less than the threshold Tl (FM~D - FDAD @Tl), the circuit 33 outputs data (data MPFM of the frame prediction mode in the motion prediction) indicating that frame-by-frame motion prediction is more eficient.
One oP these prediction mode data MPFM/MPFD is transmitted to the frame memory group fitted with the motion compensator 20, while being simultaneously supplied to sel~ctor 2~.
The sel~ctor 24 selectively outputs, responsive to prediction mode data MPFMVMPF~ from frame/field mode decision circuit 33, one Or the data FMMV of the ~rame-to~rame motion vector supplied from the ~rame motion detection circuit 22 and the data FDMV of the field~to-field motion vector supplied from the field motion detection circuit 21. That is, the selector 2 J~ ~ ~ 3 selects and outputs the motion vector data F~MV ~rom the field mo~ion detcction circuit 21 when the pr~diction mode data is the data MPFD indicating the field prediction mcde data9 while selecting and outputting the motion vector data FMMV from the frame motion detectlon circuit 22 when the prediction mode data is the da~a MPFM indicating ~he ~rame prediction mode data. The motion vector data FMMV/FDMV, as selected by selector 24, is transmitted to frame memory ~itt~d with motion compensator 20.
The ~rame memory fitted with motion compensator 20 is now able to e~ect ~rame-by-frame or field-by-field rnotion compensation on the basis of the predic~ion mode data MPFMVMPFD and motion vector data FM~V/F~V.
The frame/field mode decision circuit 33 is also supplied with picture data as read out from the frame memory group 10 on the macro-block basis. The frame/field mode decision circuit 33 also performs the operation oe produc;ng a differential picture ~rom the predi~tion mode dat~ MPFM/MPFD, motion vector data FMMV/FD~V and the picture data rrom the frame memory group 10 and selects the processing mode ~or bl~ck formation for orthcgonal transeorm (the above-mentioned frame processing mode/field processing mode) most suitable for the picture which is outputted from the frame memory group 10 and prccessed by DC~ by the DCT
circuit l3. I the current picture is the I-picture or I-~rame, data of th picture of the frame memory group 10, that is the original picture, are used.
I the difference between the difference EFM found on the frame-by-~rame basis and the di~erence FFD found on the field-by-~ield basis, using the equ~tions (1) and (2)7 iS ~ound to be larger than a threshold T2 (EFM - EFD > T2), the frame/field mode deoision circuit 33 outputs data indicating that the DCT by the DCT circuit 13 be performed on the ~ield-by~field basis (data MDFD for the field processing mode in the block forming operation for orthogonal trans~orm). Conversely, if the difference between the diPferences EFM and EFD is equal to or less than the threshold T2 (~F~ - ~FD > T2), the frame/~ield mode decision circuit 33 ou~puts data indicating that the DCT by the ~CT
circuit 13 be per~ormed on the frame-by-frame basis (data MDFM
for the framc processing mode in the block forming operation for orthogonal transform).
The output of the frame processing mode data MDFM or the 7 ~
frame processillg mode data MDFD from the frame/field mode decision circuit 33 is responsive to the first limitation mode or the second limitation mode from the limitation mode decision circuit 34 (EN1/EN2).
The limitation mode decision circuit 34 decides, using the macro-block based picture data as read from the ~rame memory group 10, which of the first and s~cond limitation modes is more efficient for encoding, and outputs the encoding mode d2ta ENl or EN2 depending on the results of decision. Specifically, the limitation mcde decision circuit 34 calculates the sum of the dif~erences of absolute values of the pixels between the ~dd ~ields (first fields) and the even fields (second fields) of the ~rames to output the limitatiQn mode data ENl indicating that the eneoding under the sesond limitation mode is more e~ficient if the sum value is less than a certain threshold TOg that is if the pic~ure experiences little motion, while outputting the limitation mode data EN2 indicating that the encoding under the first limitation mode is more efficicnt if the sum value is larger than the threshold TO, that is if the pic~ure experiences acute motion.
Meanwhile, the d~cision by the limitation mode d~cision circuit 34 may also be given using the motion vector data FDMV
from the field motion de~ection cir~uit 21~ That is, it is also possible to select the second limitation mode i the motion vec~or data FDMV between the odd field and the e~en field is less ~han a certain threshold tO and to select the first limitation ~ e if the data F~MY is larger than the threshold tO.
By the limitation mode data EN1/EN2 being tr~nsmitted from the limitation mode decision circuit 34 to the ~rame/field mcde decision circuit 33, the ~rame pro~essing mode data ~DFM or the field processing mode data MDFD responsive to the encoding msde data EN1/EN2 is outputted from the frame/field ~3de decision circuit 33.
That is, if the limitation mode data from the limitation mode decision circuit 34 is the data ENl indicating the second limitation mcde, the frame/field mQde decision cir¢uit 33 performs an operation of adaptively changing over the frame processing mode to the ~ield processing mode or vice versa for each macro-block in one ~rame. Consequently, the ~rame/ield m~de decision circuit 33 outputs the adaptively changed over .

2 ~ 7 ~
frame processing mode data MDFM or ~ie~d processing mode data MDFD.
Conversely, if the limitation mQde da~a from the limi~ation mode decision circuit 34 is the data ~2 indicating the first limitation mode, the frame/field mode dccision circuit 33 performs the operation of division into b]ocks of all of the macro-blocks in one ~rame for orthogonal transform in accordance with the field processing mode. Cons~quently, the field processing mode data MDFD is outputted ~r~m the ~rame/field mode decision circuit 33.
The ~rame/field orthogonal transform block-forming ~ode data MDFM/MDFD9 outputted from the frame/fiPld mode decision circuit 33~ and the limitation mode data EN1/EN2 from the limitation mcde decision circuit 34, are transmitted to the address generator 35 and to the ~ramc memory group fitted with the motion compensator 20. Meanwhile, the pr~cessing mode data MDFM/MDFD, encoding mode data EN1/EN2 and the motion vector data FMMV/FDMY are also transmitted to the above-mentioned variable length encoding circuit 15.
The address generator 35 controls the frame memory group 10 to output picture data of the macro-blocks divided into blocks in accordance with the processing mode data MDFMVMDFD and the limitation mode data EN1/EN2 on the macro-block basis.
Specifically, address genera~or 35 controls the ~rame memory group 10 to output macro-blocks divi~ed into blocks in accordance with the block-forming mode for orthogonal transform (data MDFM/MDFD) eor odd cycles i~ the limitation ~ e data EN1/EN2 is the data EN1 indicating the s~cond limitation mode, while controlling the ~rame memory group 10 to output macro-blocks divided into blocks in accordance with the field processing mode (data MDFD~ ~or odd and even cycles if the limitation mode data ENl/EN2 is the data EN2 ind;cating the first limitation mode.
In other words, i~, with the second limitation mode having been seleoted and the limitation mode data EN1 being supplied to address generator 35, the processing mode data is MDFM indicating the ~rame-by-~rame DCT, address generator 35 controls the ~rame memory 10 to output macro~blocks in which even and odd fields are alternately scanned, that is framc-bas~d macro-blocks each combined from odd and even fields, as shown in Fig.3. That is, address generator 34 contro]s the frame memory group 10 to divide 2 ~
a macro-block having lines 1 to 16 into lines 1 to 8 and lines 9 to 16 and to output ~our o~ 8 8 b]ocks, as shown in F;g.3.
On the other hand, if, with the second limitation mKde having been selected and the limitation mcde data ENl b~ing supplied to address generator 35, the processing mode data is MDFD indicating the ~ield-by-field ~CT, address generator 35 controls the fr~ne memory 10 to output macro-blocks in which even and odd fields are scanned s~parately~ that is separate ~ield-~ased macro-blocks for even and odd rields, as shown in Fig.4.
That is, address generator 34 divides the lines 1 to 16 into lines 1, 3, 5~ 7, 9, 13 and 15 ~lines of the odd fields or the first fields) and lines 2, 4, 6, 8, 10, 12, 14 and 16 (lines of the even ~ields or the second fields), as shown in Fig.4, and to output two 8 * 8 blocks for each o~ these cdd ~ields and even fields.
On the other hand, i~, with the ~irst limitation mode having been selected and the limitation mode data EN2 being supplied to address generator 35, address generator 35 controls the fr~me memory group 10 to output macro-blocks divid~d into blocks in accordance with the field processing mode for the odd and even cycles, as described previously. That is, if the first limitation mode is selected, address controller 35 controls the frame memory group 10 so that two 8 * 8 blocks (only lumin ~ ce components, as described subsequen~ly) are outputted at all times. Specifically9 address generator 35 ~ontrols the frame memory group 10 so that the two 8 * 8 block macro-block) is outputted in an amount corresponding to one frame (one picture) only for odd fields during the odd cycles, while controlling the ~rame memory group 10 so that the two 8 * 8 block (macro-block) is outputted in an amoun~ corresponding to one fra~e (one picture) only for even fields during thc even cycles.
The picture data outputted ~rom the frame memory group 10 controlled by address generator 35 is processed with ~cr by the DCT circuit 13, as described above. If, for example, the s~cond limitation mode and the frame processing mode are selected, DCT
circuit 13 eefectuates DCT on unit blocks of 8 * 8 pixels as shown in Fig.3. If, for example, the second limitation mode and ~he field processing mode are selec-ted, DC~ circuit 13 e~fectuates DCT on unit bl~cks of 8 * 8 pixels as shown in Fig.~.
If the first limitation mode is selected, DCT circuit 13 effectuates DCT on the 8 * 8 pixel blocks only for odd ficlds during the odd cycles and on the 8 * 8 pixel blocks only for even fields during the even cycles.
Besides, the prediction mode da~a MPFM/MPFD and processing mode data MDFM/MDFD from the frame/field mode decision circuit 33, motion vector data FMMV/F~YV as sc]~cted by selector 24 and the limitation mode data ENl/EN2 from the limitation mode decision circuit 34 are also supplied to the frame memory fitted with the motion compensator 20. Thus the ~rame memory fitted with the motion ccmpensator 20 is not only responsive to the prediction mode data MPFMVMPFD o~ motion prediction, processing mode data MDFM/MDFD of DCr processing and to the encoding mode data EN1/EN2, but also effectuates motion compensation with the aid of the motion vector da~a FMMV/FDMV.
For the second limitation mode and the frame processing mode, motion detection of the P and B ~rames may be n~de from the forward and backward frames, as shown in Fig.29. Thus, in the DCT circuit 13, a diferential picture from the prediction-coded picture is processed with DCT by unit blocks of 8 * 8 pixels.
In Fig.29, the forward, current and backward frames are shown, with the arrow indicating thP motion v~ctor ~nd MB macro-blocks.
For the first limitation mode and the field processing mode, motion detection of the P and B frames may be made from the odd and even ields (first and second fie]ds~ of the forward and back~ard frames for each o~ the ~dd fields and even fields, as shown in Fig.30. In Fig.30~ odd and eve~ fields of the forward, current and backward rames are shown, with the arrow mark indicating the mation vector and MB macro-blocks.
For the ~irst limitation mode and the ~ield processing m~de, motion predic~ion of the odd and even fields of the macro-blocks is pcr~ormed ~rom the odd and even fields of the ~orward and backward ~rames, as shown in Fig.31. Motion prediction between the fields in e~ch frame is also made. Consequently, a differential picture from the prediction-coded picture is processed with DCT by the DCT circuit 13 by the unit 8 * 8 pixel blocks. In Fig.31, odd and even fields of the forward, current and backward frames are shown, with the arrow mark indicating the motion vector and MB macro-blocks.
With the above-described high efficiency encoding device of ~he embodiment 3, high efeiciency encoding is achieved by ~Q~t~ 7~

changing over between encoding without intra-rrame interfield prediction and encoding with intra-frame inter~ield prediction depending on the first and second limitation modes, that is on the degree of motion in the picture. Above all, the first limitation mode is effective for a framc with significant movement.
Meanwhile~ with the encoding device of the embodiment 3, the present code is realized by adding extension bits to the conventional macro-block t ~ e for interchangeability with the oonventional system.
That is, in the embcdiment 3, the macro-block type has three predictions, namely forward prediction, backward prediction and bidirectional predietion. Since prediction from odd-numbered ~ields o~ the previous frame ~nd prediction-coded ~ields of the previous ~ield are possible with ~he forward prediction, the present code is realized by appending ex~ension bits use~ul for r~cognizing which o~ these predictions is used. Since there are two predictions, it suffices to add one bit for one direction, that is for the forward or bac~ward direction. For Pxample, if the prediction is ~orward or baekward pre~iction ~nd from the odd-numkered fields, it suffi~es to add a code 1 as an ex~ension bit to the conventional macro-bit type. Similarly, if the prediction is ~rom the even numbered field, it suffices to add a code O as an extension bit to the conventional m2cro-bit type.
For bidirectional prediction, both extension bits are appended for both ~orward and backward predictions.
Meanwhile, for a frame prediction m~de, no extension bits are a ~ nded, so that the bit string is o~ the same style as the conventional bit stream (MPEG).
The akove applies to the P-~rame as well.
In the embodiment 3, similarly to the preceding embodiment, the macro-block type has three predictions, namely forward prediction, backward prediction and bidirectional prediction for e.g. the B-frame. If the prediction is the forward prediction with the ~ield prediction mode, it is necessary to append extension bits to the macro-block for allowing to recognize if, when the prediction is for forward pr~diction and the field prediction mode, the prediction is that ~rom an odd-numbered field, an even-numbered field or ~rom an odd-numbered field of the current frame. That is, since there is a prediction fr~m the ., 2 0 ~ ~J 11~ 7 ~3 own frame in the forward field prediction mode, one or two bits are requircd for expressing three predictions including odd- and even-numbered predictions by ext~ns;on bits. On the other hand, for backward field prediction l~ode~ since there are two predictions, that is prediotion for odd and even numbers, 1 extension bit is required.
I~ the mode is the frame prediction n~e, no extension bits are annexed, so that the bit string is o~ the same style as the conventional bit stream (MPEG). For bidirectional prediction, both extension bits are annex~d for forward or backward prediction.
The above applies to he P-frame as well.
As a modificationg the number of the extension bits can be reduced to one for forward prediction. ~hat is, in the even-numbered cycle ~or the ~ield prediction mode, the number of bits n~y be reduced to two and the forward prediction mode may be transmitted by one extension bit by abolishing prediction from the even-num~ered field furthest removed bcth temporally and spatially, shown by chain-dotted line, as in the embodiment shown in Fig.16.
Fig.32 shows a typical arrangement o~ an encoding device o~
the second embcdiment 3. In Fig.32, the same components as those shown in Fig.20 are denoted by the same re~erence numerals and detailed description thereof is cmitted for simplicity.
The arrangement o~ the second e ~ iment 3 is a 3-pass encoding devi~e n which three operations are carried out for processing each frame.
That is, an operaticn by the above-d~3scribed first limitat;on mode by a fixed quantization width with intra-frame interield prediction is carried out ~or the ~irst pass, and an operation by the above-described second limitation mode by a ~ixed quantization width without intra-field interframe predic~ion is carried out ~or the second pass. The operation of the first and second passes in which a smaller number of bits have occurred is selected for the third pass which is carried out with a controlled quantization width.
~ n the second embodiment 3, a macro-block unit 55, a changeover switch 57, a ~ield block-~orming transform circuit 56 and another changeover switch 58 are conn~cted to a ~ownstream side of the ~rame memory group 10. Picture data ~rom the frame memory group 10 is transmitt~d to a moltion detection circuit 51 effeetuating ~ame and ~ield motion detcction. An output of the motion detection circuit 51 is transmitted to a pr~cessing mode decision circuit for selecting the fr~mc/~ield modes for motion detection and block division for orthogonal transform 52, frame memory group 20 and the variable length encoding cireuit 15.
Output mode data from the processin~ mode decision circuit 52 is transmitted to the frame mcmory group 20 and the variable length encoding circuit 15. 0~ these data, field processing mode data are supplied to one input terminal of a two-input ~ND gate 53. An output o~ the changeover switch 59, which is changed over depending on the pas numbers 1 to 3 is supplied to the other input terminal of the AND gate 53. An output terminal of the 2-input AND gate is connected to movable terminals of the changeover switches 57, 58.
Data of the number of the produced bits is outputted from the variable length encoding circuit 15 and transmitt~d to a selecting circuit 60 for selecting one of the first and second processing modes with smaller number of produced bits based on the data o~ the number o~ the produced bits (circuit for deciding whether or not there is intra-frame interfield prediction between the fields in one frame~. The stored volume data from buffer 16 is transmitted to the variable length encoding circuit 15 and to one o the ~ixed terminals of the changeover switch 61. Fixed values o~ ~he first and second passes are supplied to the other ~ixed terminal o the changeover swit~h 61.
In the above-described second embodiment 3, the picture entered to terminal 1 is temporarily stored in the ~rame memory group 10. Frame or field data are fetched as required from the frame memory 10 and, using these picture data, the motion vec~or is found by the motion detector 51. The processing mode decision circuit 52 gives a decision of the field/~rame mode for each macro-block from the motion prediction residues fr~n the motion detector 51. The macro-block unit 55, connected to the downstream stage of the frame memory group 10, receives the inPormation for the first, second and third passes, that is the in~ormation oP the presence/absence of intra-frame interfield prediction which is the above-mentioned second or first limitation mode, via changeover swit~h 59. If the macro-block unit 55 has received the first limitation mode information, it transmits only the block of the odd f;eld (first fic]d) and subsequently transmits the even field (sccond field)~ while turning o~ the blo¢k division of th~ frame processing mode.
The picture data the macro-blocks of which are set to the frame prccessing mode based on the information Oe the second limitation modc in the macro-block unit 55 is set to the block of the frame processing mode in the field block division conversion circuit 58. 1 bit is added to the selectGd mode information for each frame.
The block dîagram of Fig.33 shows a decoder for picture signals. The third high ef~iciency decGding device includes inverse variable length encoding means for receiving and decoding encoded data of $he playback picture and the h~ader information including detected motion vector information, processing ~ode information and limita$ion m~de information and outputting the detected motion vector information, processing mode information and limitation mode information simultanecusly with dec~ded picture da~a, address generating means ~or calculating an address increment value at a frame buffer storing the d~coded picture data fr~n the limitation mode inrormation, finding a l~ading address of each macrorblock and according the leading address ltO
the frame bu~er, and motion eompensating m~ans for supplying the rela~ive addresses of the macro-blocks other ~han the leading address, accessing lthe data, receiving the detected motion vector in~ormaltion, pr~cessing mode information and limitation mode information, executing motion cc~ nsation in association with the n~de in~ormaltion, and transmitlting motion-compensated pictllre signals to the frame buf~er.
That is, the high e~iciency decoding device of ~he present embodiment is made up of an inverse variable length enccding circuit ~1 for receiving and decoding encoded picture data and the header informa~ion including the detec~ed mntion vector information, block-forming mode information (processing m~de information) and limitation mode information (limitation mode data) and outputting the detected motion vector information, prediction mode information, processing mode information and limitation mode information of the header information simultaneously with the decoded picture data, address generators 81, 82 and 83 f'or calculating address increment values at frame buffers 61, 62 and 64 storing picture decoding data f'rom the 7 ~
limitation mode data for finding the ~ading address o~ each macro-bloc~ and supplying the leading address to the ~rame buf~ers 61, 82 and 64, and motion compensation circuits 59, 60, 63, 65 and 66 or supplying the relative addresses o~ the macro-blocks other than the leading address to the frame buffers 61, 62 and 64 for data accessing~ receiving the detect~d motion vector information, prediction mode information, processing mode information and limitation mode inrormation, executing prediction between the motion compensated frames or fields in ass~ciation with the mode information and transmitting the motion-ccmpensated picture in~ormation to the rame bu~ers 61, 62 and 64.
Referring to Fig.333 da~a encoded by the high efficiency encoding device of the embcdiment 3 are transiently recorded on a storage medium, such as CD. The encoded data reproduced form CD is supplied via input terminal 50 to an inverse variable length encGding circuit 51 where thc header in~ormation etc. are dec~ded from one sequence to another, from one frame group to another and from one frame to another. The header inPormation is decoded from one slice (macro-block group3 to another. The quantization width is included in the slide header. The macro-block type indicating the macro-bl~ck addresses9 frame/field prediction mode, frame/field pr~cessing m~de~ encoding mode data and the decQding system is decoded from one macro-block to another. The quantization width is decoded at the time of the updating.
Meanwhile, i~ the block formation in a macro-block is of the fr~me processing mode, the macro-block in its entirety is decoded ~or the odd-numbered cycle, while no data is decodcd for ~n even-numbered cycle. I the block ~ormation in a macro~block is of the ~ield processing mode5 only the block including odd numbered ~ields in the macro-block is d~coded for the odd-numbered cycle, while the block including even-number~d ~ields is decoded for the even-numbered cycles.
The picture ineormation is decoded by a de~uan~izer 53 performing an operation which is the reverse o quantization and an inverse DCT circuit 54 per~orming an inverse DCT operation, and a decision is given by the macro-block type as to whether or not the inform~tion is a differential information. ~epending on the results Oe decision, a mode switch 57 for c~angeover between addition or non-addition to the reference picture (or between non-intra/intra of ~EG c~ding) by an additive node 56. The decoded picture is entered to frame burfers 64 or 61 for the I~or P-frame (alternately each time the I-~rame or the P-fr~me is processed) and to ~rame bu~f~r 62 ~or the B-~rame. Each ~rame buffer is ~omposed o~ two field burrcrs and the odd/even field picture is stored separately in each o~ these field buffers.
Recording in the frame buffers is controlled by changing w er a switch 58.
The addresses recorded in the framc buf~ers 61, 62 and 64 are supplied from address generators 81, 82 and 83. The address generators 81, 82 and 83 calculate an address increment value in the frame buffers 61, 62 and 64 fr~m the encoding mode data in the header information of the macro block to find the leading address of each macro-block.
The quantization width data is stored in a one-field memory 52. The quantization width data is supplied to dequantizer 53 via switch 55 changed over depending on the output o~ the inverse variable length encoding circuit 51. Since only the macro-block processed with the ~ield processing mode is decoded ~or the even-numbered cycles, the macro-block address decoded for each macro-block, the macro-block type and the motion vector requir~d by the prediction system indicated thereby are de~oded and the differential pic~ure from the reference field is added to the motion-compensated picture to produce the playback picture.
On the other hand, data of the frame bufers 64, 62 and 61 is motion-compensated by the motion compensation circuits 65, 66, 59, ~0 and 63. At this time, the respective motion compensation circuits change over the frame motion compensation/field motion compensation (~rame/~ield) depending on the block-forming mode at the time of the DCT operation.
These motion-compensated pictures are transmitted to respective fixed terminals of changenver switches 67, 68 and 71.
These changeover switches 67, 68, 71 are changed over ~or taking out the referen~e rield or ~rame indicated by the decoding system of the macrorblock decoding type. The changeover switch 71 is supplied wi~h an output of switch 67 and a signal which is outputs of the changeover switches 67 and 68 added t~gether by additive node 69 and subsequently halved by a divider 70. An outpu~ of switch 71 is supplied to switch 57.
Outputs of ~rame buffers 64, 61 ad 62 are supplied via 2 0 ~ ~ 1 7 ~
changeover switch 72 to a display 73. The display 71 is supplied with the outputs of the changeover swilch 72 which are changed over in the s~quence o~ playback pictures instead o~ in the decoding sequence for pr~dueing an output picture.
Meanwhile, with the resent ~mbodiment, since only the macro-block processed with the field processing m~de is decoded during the even cycle~ it is necessary ~o know a n~cro-block address.
There are two meth~ds for grasping the macro~block address~ One is to transmit the address Oe the macro-block ~or each macro-block o~ the even cycle as described previously. The other method is to store the in~ormation of the one-~ield ~ield processing mod~/frame processing mode information during the odd-numbered cycle and to calculate the address of the macro-block in the field processing mode ~rom the string o~ each processing mode. The former method has an adv~ntage that it is unnecessary to add memories, whereas ~he latter method has an advantage that the transmitted in~ormation is not increased. The same may be said of the quantization width if the quantization width is transmitted for each macro-block wi~hout storing one-field data during the above-mentioned odd-numbered cycle.

*** FIFTH EMBODIMENT ***
m e advantages o ~he fifth encoding device shown in Fig.34 are explained. In this figure, the blor~ks denoted by the same numerals as those of Figs.1 and 5 have lthe same ~unction.
There~ore, only the blocks denot~d by diferen~ numerals from those of Figs.1 and 5 are explained.
The high efficiency encoding deYice o~ Fig.34 includes, in addition to the blocks denoted by the same numerals as those of the high efficiency en~Gding device shown in Figs.1 and S, an encoding mode decision circuit 34(c) and a selector 24, as limitation mode selecting means, for selecting the ~irst limitation m~de of inhibiting the encoding of the entire macro-blocks in one frame by the above-mentioned frame processing mode or the second limitation mode of i~hibiting prediction of the even field o the current frame from the odd field of the same frame in the entire macro~blocks in one frame, whichever is more efficient, sel0cting only the first limitation mode for the bidirectional predicted frame (B-frame) and inhibiting prediction o~ the even field from the odd field o~ the B-picture, and an 2 ~ 7 ~
address generator 35(c), as address generating means, ~or controlling the frame memory group to output odd field c~mponents of the entire macro blocks and to subsequently output even field components of the entire macro-blocks in case of selection of the first limitation mode ~or one rame, as well as to output the macro-blocks sequentially by one slice at a time on the basis of the frame c~mposed o~ the odd and even ~ield components of the entire macro-blocks in case of selection o~ the second limitation mode.
It is assumed that, with the encoding device of the e~bodiment 59 the picture sequence in the display time is BOo, BOe, Ilo, Ile, B2c, B2e9 P30, P3e, B40, B4e, P50, P5e, ....
as shown in Fig.35.
The code sequence or decoding sequence o~ the present emb~diment is Ilo, Ile, BOo, BOe, P30, P3e, B20, B2e, P50, P5e, B40, B4e.
Meanw~ile, for decoding the codes by the second encoding device using a deesding device, playback is rendered possible by having three frame buffers~ that is six fields, at the maximum.
The operation o~ the decoding device having 3-frame frame buf~ers A, B and C is explained by referring to Fig.36. For using the frame buffers, the decoded pictures of the I-frames or P-frames are stored by alternately changing over between the frame bu~ers A and B1 Display is made of the contents of the frame buffer opposite to that used for storage of the current ~rame and in t~e sequence of odd and even fields. In other words, ~or storing the picture in ~rame buf~er A, the contents of the frame buffer B are displayed, whereas, for storing the picture in frame bufPer B, the ~ontents oP the ~rame buffer A are displayed. Two ~rames need to be stored in the frame buffers so as to be used as reference ~rames for motion compensation of the B- or P-frames to be decoded subsequently.
For deooding the B-fra~e, the decoded picture is stored in frame bufer C. Display is made of the contents of the frame buffer C in the sequence of the odd fields and the even fields.
If decoding is performed according to such rule, the entire components of IlO and the d~coded picture o~ the even components o~ the macro-blocks of the ~rame-based block-forming frame-based prediction mode o~ Ile are stored at time l, while motion comp~nsation is made at time le by referring to the contents of 7 ~
the frame bufer A at time le and the decoded picture of the even field componen~s of the macro-blocks of the field-based block forming field-based prediction mode of Ile is stored at time le in frame buffer A.
At time 10, and at time le, the odd and even components of frame buf~er B are displayed, respectively. In the absence of the previous codes, the contents of thc frame buffer B at this time point become indefinite.
At this time 20, the pictures of the frame buffers A and B
are motion-compensated, and the entire components of BOo and the decoded picture of the even field components of the macro-blocks of the frame-based block forming frame-based predi~tion mode of BOo are stored. At time 2e, the pictures of the frame buffers A, B and C are motion-compensated, and the dec~ded picture of the even field-based c~mponents of the macro-blocks of the rame-based bloek orming frame-based prediction mode of ~Oe are stored.
At time 2e, the pictures of the framc buffers A, B and C are motion-compensated, and the decoded picture of the even fiel~
components of the macro-blocks of the frame-based block forming-based frame prediction mode of BOe are stored in the frame buf~er C.
At time 20 and at time 2e, the odd ~nponents ~nd even components of the ~rame bu~er C are displayed, respectively.
~ hen BOe is displayed, the odd cycle of the B-frame is being decoded9 such that koth the components of BOo and bQe are contained. Therefor~, the picture components of BOe, decoded at this time, need to be stored for being displayed temporally posteriorly.
When BOe is displayed, the even cycle of the B-Prame is being dec~ded, ~ld the remaining components of BOe which has not been decoded during the odd cycl~ are decoded. Consequen~ly, since it is neccssary to m~e motion cvmpensation of the pictures from BOo to BOe, the picture o BOo needs to be stored.
7herefore, a one-rame frame buffer becomes necessary for the B-frame, such that a three frame bu~fer is required for decoding the codes ~ormed by the second encoding device.
With the fifth encoding device, as shown in Fig.37, the encoding mode o~ the B-frame is carried out only by dividing the field into blocks or field prediction, while prediction of an 2 0 .~ ~ ~ 7 3 even field from an odd field of the B-~rame is inhibited. Thus, as shown in Fig.35, only BOo is de ~ cd at time 20 and simultaneously BOo is display~d. This picture nced not be stored because it is not used ~or subsequent motion compensation.
At time 2e, only BOe is decoded and simultaneously BOe is displayed, so that BOe need not be storedO Cons~quently, the decoding devicc ~or decoding the codes prepared by the fifth encoding device need not be provided with the frame buffer C.
Such c~des may be decoded by a decoding device having only a 2-~rame buffer, that is a four-field buffer, as shown in Fig~38s ~or enabling the size and costs o~ the decoding device to be reduc~d.
At time 30, the picture of frame buffer A is motion-compensated, and the entire c~mponents of P30 and the decoded picture of the even field components o~ the macro-blocks of the frame~based block forming frame-based prediction mode of P3e are stor~d in frame bu~fer B. At time 3e, motion ~ ensation is made by referring to the contents of the ~rame buffers A and B
and the decoded picture o~ the even field components o the macrorblocks of the ~ra~e-based block forming field-based prediction mode of P3e are stored in frame buf~er B.
At time 3e, the pictures of frame buffers A and B are motion-compensated, and the decoded picture of the even field ccmponents of the macro-blocks o~ the ~;eld-based block ~orming field-based prediction mode o P3e are stored in frame buffer B.
At time 3e ~nd a~ time 3e, odd ~ onents and even components o~ th~ frame buffer A are displayed, respectively.
Subsequently, ~e decoding and display proceed in a similar manner.
I~ the GOP ssquence is BOo, BOe, Blo, Ble, I20, ~e, B30, B3e, B~o, ~ e, P50 P5e, as shown in Fig.39, that is, i~ de~oding is performed so that there are two B-frames between I and P ~rames or between two B fr~mes, decoding may be made by the same de~oding device, if the above-described d~coding sequence is used, as shown in Fig.38.
The same may be said o a case wherein there are more than two B ~rames between frames or between two P frames.

*** SIXTH EMBQDIMENT **~
The advantages Or the sixth encoding device, arranged as , 2~J~-7~
shown in Fig.40, are explained. In this figure, the b]ocks denoted by the s~mc numcrals as those of Figs.l ~nd 5 have the same function. Therefore, only the blocks denoted by different numerals from those o~ Figs.l and 5 are explained.
The high efficiency encoding dcvice o~ Fig.40 includes, in addition to the blocks denot~d by the same numerals as those of the high e~iciency encoding device shown in Figs.l and 5, an encoding mode decision circuit 34(d) and a selector 24, as limitation mode selecting means, for selecting the first limitation mode of inhibiting the encoding of the entire macro-blocks in one frame by the above-mentioned frame processing mGde in the entire macro-loc~s in one ~rame, or ~he se~ond limitation m~de of inhibiting prediction of the even field of the current frame from the odd field of the same ~rame, whichever is ~ore e~ficient, selecting only the first limitation mode ~or the bidirectional predicted frame (B-~rame), inhibiting prediction o the even field ~rom the odd field of the B-picture, and i~hibiting prediction from ~n odd field of a frame which is to be a reference frame for forward prediction for the B-picture, and an address generator 35(d), as address ~enerating means, for controlling the frame memory group to output od~ field ~omponents of the entire macro-blocks and to subsequently output even field components of the entire macro-blocks in case o~ selection of the first limitation mode for one frame, as well as to output the macro-blocks sequentially by one slice at a time on the basis of the frame composed o~ the odd and even field components o~ the entire macro-blocks in case of selection o~ the second limitation mode.
It is assumed that, in the present embodiment, the display time sequence is BOo, BOe, Blo, Ble, I20, I2e,. B30~ B3e~ B40, B4e, P50 P5e, ..., as shown in Fig.41.
The coding sequence with the encoding device o~ the present emb~diment is I20, I2e, BOo, BOe, Blo, Ble, P50 P5e, B30, B3e, B40, B4e....
With the sixth enco~ing device, shown in Fig.40, forward prediction from an odd field o~ the B-frame, as used in the fi~th encoding device, is inhibited, as shown in ~ig.42.
Consequently, the picture necessary for prediction may be dccoded by a decoding device having a bu~fer provided with a ~rarne (two ~ields) for backward prediction and a field for 2 ~ 7 ~
forward prediction9 that is a buffer ror three ~ields, as shown in Fig.43.
This will ~e explained by referring to Fig.41.
In using the frame bufer, decod~d pictures of I- or P-frames are stored by ch3nging over the f;eld buffers in the sequence of the riel~ buffers A, B, C, A, B, C, from field to field. When storage is initiated at the field buffer C, the contents of the field buffer A are disp1ayed0 ~imilarly, when storage is initiated at the field buffer A, the contents of the ~ield buffer B are displayed, whereas, when storage is initiated at the field buffer B~ the contents of the field buffer C are displayed. For decoding the B-frame, the encoded picture is not stored, but displayed instantly.
If the decoding is continued in this sequence, the following operation is incurred. At Io, since storage is initiated at the ield buffer Ag ~he ~on~ents of the field bufPer B are displayed.
At Ie, since storage is initiated at the field buffer B, the contents o the field buffer C ar~ displayed. If there is no previously deccded picture, the display ~ontents be~ome indefinite.
At time Io, I20 in its entirety and even components of the macro-block of the fr~mc-based block-forming frame-based prediction mode of I2e are decoded and stor~d in field buffers A
~nd B, respectively.
At time le, by I~Dtion-compensating the picture of the field buffer A, even components of the macro-block of the field-based block-forming ield-based prediction mode o I2e are decoded and stored in ield bufer B.
At time 20, since the picture to be encoded is the B-rrame~
the pictures of the field buffers A, B and C are motion-compensated and BOo dec~ded. The deccded pictures are not stored but displayed instantly. The same may be said of BOe, Blo and Ble at time 2e, time 30 and a~ time 3e, respectively.
At time 40, since storage is initiated at field buffer C, ~he contents of the field buffer A are displayed. At time 4e, since storage is initiate~ at ficld bufrer A, the contents of the rield buePer ~ are displayed.
A~ time 40, the pictures of the field buffers A and B are motion- ~ npensated and P30 in its entirety and even components Oe the macro-block of the ~rame~based block~forming ~rame-based ~ 7 3 prediction mode of P3e are decoded and stored in field bur~ers C
and A.
At time 4e, by motion-compensating the pictures of the field buffers B, C and A, even components a~ the macro-block of the field-based block-forming field-bascd prediction mode of P3e are decoded and stored in ~ield bu~cr A.
It is seen from above that, with the sixth encoding device, decoding is made possible by a framc buffer -~or three fields, that is one and a half frame, so that the buffer employed in the decoding device may be reduced in capacity, as shown in Fig.43 and the decoding d~vice may be r~duced in size ~nd costs.
The output bit stream of the transmission buffer 16, shown in Fig.1, is multiplexed with encoded audio signals, synchronization signals etc., added tu by error correction code data and modulated in a predetermined ma~ner before being recorded on a recordin~ medium, such as an optical disc9 tape or a semioonductor memoryj by m~ans of a laser beam.
The bit stream is inputted to the decoder on a transmission me~ium, such as an optical disc. The playback data reproduced from the transmission medium is demodulated and corrected for errors. If the bit stream is multipl~xed with audio signals, synchronization signals etc., it is separated from these signals.
~: Although the bit stream outputted from the encoder is recorded herein on the optical disc, it may also be transmitted to a transmission channel for ISDN or satellite communication.
With the above-described first and second high effici~ncy encoding devi~es according to the present invention, ~ield-by-field moving pictures having little motion and/or acute motion may be processed e~Piciently on the ~ield-by-field or frame-by-frame basis. A~ove all, with the second high efficiency encoding device, it is possible to predict an even-numbered field from an odd-numbered field of the current frame and to select the encoding with higher efficiency. Therefore, a high-quality moving picture may be reproduced by the high efficiency decoding device of the present invention despite the small volume of the transmitted information.
With the above-described third and four~h high efficiency encoding deviccs according to the present invention, since the frame prediction inhibiting mode is changcd over to the mode of inhibiting prediction of the even-numbered field from the odd-2 ~ 7 .~
numbered field o~ the current frame or vice versa on the frame-by-frame or slice-by-slice basis~ address generating means for encoding and address generating means ~or decoding may be simplified to enable the hardware to bc reduced in size.
With the fifth high efficiency encoding device, since the ~-frame is processed in its entirety on the ~ield-by-field basisg and the mode of inhibiting prediction of the even-numbered field from the odd-numbered ~ield of the current frame is changed over on the rame-by-frame or slice-by-slice basis, it becomes possible to use the frame buffer having two frames or ~our fields for the decoding device.
With ~he sixth high efficiency enooding device, ;t becomes possible to use ~he frame buffer having one and a half frame, ~hat is three fields.
The first to sixth dec~ding devices according to the present invention may be used ;n conjunction with the first to sixth encoding devices for realizing high ef~iciency decoding.
Since compressed data encoded by the above-described high effioiency encoding devices are recorded on the recording medium ac~nrding to the present invention, it be~cmes possible to record more picture data, that is picture data of longer time duration, on the recording medium.

,, :
,

Claims (57)

1. A picture signal encoding method for encoding data representing interlace scanned pictures, comprising the step of:
receiving interlace scanned picture data; and transforming said interlace scanned picture data by adaptively selecting frame-based DCT transformation or field-based DCT transformation.

2. A picture signal encoding method according to claim 1, further comprising the step of dividing said interlace scanned data into macro-blocks, and selecting field-based or frame-based DCT transformation for each macro-block.

3. A picture signal encoding method according to claim 1, further comprising storing a preceding frame of interlace scanned data; providing compensation for movement between received interlace scanned picture data and the preceding frame of interlace scanned data to produce predictive picture data; and subtracting said predictive picture data from said received interlace scanned picture data to produce picture data for transformation.

4. A picture signal encoding method according to claim 3, wherein the step of providing compensation comprises adaptively selecting frame-based movement compensation or field-based movement compensation.

5. A picture signal encoding method according to claim 4, wherein said interlace scanned data is divided into macro-blocks and selecting frame-based or field-based movement compensation for each macro-block.

6. A picture signal encoding method for encoding interlace scanned picture data, comprising the steps of:
receiving a frame of interlace scanned picture data, said frame including first and second fields of picture data; and encoding said interlace scanned picture data by adaptively selecting a first encoding technique from which a second field of picture data is predictable from a first field of picture data in the same frame or a second encoding technique from which a second field of picture data cannot be predicted from a first field of picture data in the same frame.

7. A picture signal encoding method according to claim 6, wherein said first encoding technique includes the steps of dividing each frame into blocks of picture data derived from said first field and blocks of picture data derived from said second field; and orthogonally transforming said blocks.

8. A picture signal encoding method according to claim 6, wherein said second encoding technique includes the steps of dividing each frame into macro-blocks of picture data, each macro-block containing plural blocks; selectively changing over between field division of blocks in which the blocks of a macro-block are separated into blocks from said first field and blocks from said second field, respectively, and field non-division of blocks in which the blocks of a macro-block are separated into blocks from both said first and second fields; and orthogonally transforming the field division blocks and the field non-division blocks.

9. A picture signal encoding method according to claim 6, wherein said step of encoding includes generating head information for the encoded picture data; and further comprising adding ID data to said header information to distinguish said first encoding technique from said second encoding technique.

10. A picture signal encoding method for encoding a frame of interlace scanned picture data, with each frame having first and second fields, said method comprising the steps of:
encoding said first field to produce first encoded data;
decoding said first encoded data to produce first decoded data;
producing predictive picture data from said first decoded data; and using said predictive picture data to encode said second field to produce second encoded data.

11. A picture signal encoding method according to claim 10, wherein said predictive picture data is produced by using most recently encoded and decoded data from said first field in accordance with a picture encoding technique selected from the group consisting of intra-frame encoding and predictive picture encoding.

12. A picture signal encoding method according to claim 10, further comprising the step of dividing each field of picture data into macro-blocks; and said second field is encoded by adaptively selecting a first encoding technique for a macro-block by using a different between said second field and said predictive picture or a second encoding technique for said macro-block using only said second field.

13. A picture signal encoding method for encoding interlace scanned picture data by selectively encoding said picture data by an intra-coded picture encoding technique or a predictive coded picture encoding technique, the picture data having field intervals, said method comprising predicting a field to be encoded using said predictive coded picture encoding technique by decoding the two last fields previously encoded by a selected one of said intra-coded picture encoding techniques or said predictive coded picture encoding technique.

14. A picture signal encoding method according to claim 13 comprising generating header information for the encoded picture data and adding ID data to said header information to identify the encoded fields which are used to predict said field to be encoded.

15. A picture signal encoding method for encoding interlace scanned picture data by a bidirectionally-predictive coded picture encoding technique, comprising:
storing picture data from a previously encoded frame;
storing picture data from a following frame to be encoded; and predicting a field in a current frame to be encoded by said bidirectionally coded picture encoding technique by using stored picture data from said previously encoded frame and stored picture data from said following frame but not picture data from another field of the current frame to be encoded.

16. A picture signal encoding method according to claim 15, further comprising generating header information for the encoded picture data and adding ID data to said header information to identify which fields in said previously encoded frame and in said following frame are used to predict said field.

17. A picture signal encoding method according to claim 15, wherein said field is predicted from both fields of said previously encoded frame and both fields of said following frame.

18. A picture signal encoding method for encoding interlace scanned pictures having frame intervals formed of field comprising the steps of selectively predicting a current picture by frame-based movement compensation or field-based movement compensation of successive pictures; providing frame-based DCT
transformation of the interlace scanned pictures when frame-based movement compensation is provided; and providing field-based DCT
transformation of the interlace scanned pictures when field-based movement compensation is provided.

19. A picture signal encoding method for encoding interlace scanned picture data formed of frames having odd and even fields and in which said picture data is selectively encoded by an intra coded picture encoding technique, a predictive coded encoding technique or a bidirectionally-predictive coded picture encoding technique, said method including the step of predicting an even field of a present frame to be encoded from an odd field of the same frame when either said intra coded picture encoding technique or said predictive coded picture encoding technique is selected for encoding but not when said bidirectionally-coded picture encoding technique is selected for encoding.

20. A picture signal encoding method for encoding interlace scan picture comprising, encoding said interlace scan picture by using a fixed quantization width with interfield prediction within one frame, encoding said interlace scan picture by using a fixed quantization width without inter-field prediction within one frame, encoding said interlace scan pictures by using a controlled quantization width with or without said inter-field prediction within one frame.

21. A picture signal decoding method for encoded data representing interlace scanned pictures, comprising the steps of receiving said encoded data and inverse transforming said encoded data by adaptively selecting frame-based IDCT transformation or field-based IDCT transformation to recover decoded picture data.

22. A picture signal decoding method according to claim 21, wherein said received encoded data is divided into macro-blocks, and frame-based or field-based IDCT transformation is selected for each macro-block, said selecting is in a macro-block as a unit.

23. A picture signal decoding method according to claim 21, further comprising the steps of storing previously decoded picture data, sensing motion between a picture represented by the stored picture data and presently decoded picture data, motion compensating the stored picture data in response to the sensed motion to produce predictive picture data, and adding said predictive picture data to said presently decoded picture data to produce picture data for storing.

24. A picture signal decoding method according to claim 23, wherein the step of motion compensating comprises adaptively selecting frame-based motion compensation or field-based motion compensation.

25. A picture signal decoding method according to claim 24, wherein said encoded data is received in macro-blocks, and wherein frame-based or field-based motion compensation is selected for an entire macro-block.

26. A picture signal decoding method for encoded interlace scanned picture data comprising the steps of:
storing a previously received frame of picture data, said frame including first and second fields; and decoding said encoded picture data by adaptively selecting a first decoding technique from which a second field of picture data is predictable from a first field of decoded picture data in the same frame or a second decoding technique from which a second field of picture data cannot be predicted from a first field of decoded picture data in the same frame.

27. A picture signal decoding method according to claim 26, wherein said first decoding derives blocks of picture data to form a frame and wherein each block is formed of picture data from both said first and second fields of said frame.

28. A picture signal decoding method according to claim 26, wherein said second encoding technique forms a frame from macro-blocks of decoded picture data, each macro-block containing plural blocks, and includes the step of selectively changing over between field construction of a macro-block in which blocks from said first field and blocks from said second field are combined into a macro-block, and field non-construction in which blocks from both said first and second fields are combined into a macro-block.

29. A picture signal decoding method according to claim 26, wherein said first or second decoding technique is selected for each frame.

30. A picture signal decoding method according to claim 26, wherein the encoded picture data includes header information include ID data to identify the encoding technique of said encoded picture data and the step of selecting is responsive to said ID data.

31. A picture signal decoding method for frames of encoded interlace scanned picture data with each frame having first and second fields, said method comprising the steps of:
decoding said first field to produce first decoded data, producing predictive picture data from said first decoded data; and using said predictive picture data to decode said second field to produce decoded interlace scanned picture data.

32. A picture signal decoding method according to claim 31, wherein said predictive picture data is produced by using most recently decoded data from said first field in accordance with a picture encoding technique selected from the group consisting of intra-frame encoding and predictive picture encoding.

33. A picture signal decoding method according to claim 31, further comprising the step of constructing each field of picture data from macro-blocks; and said second field is decoded by adaptively selecting a first decoding technique for a macro-block by using a difference between said second field and said predictive picture or a second decoding technique for said macro-block using only said second field.

34. A picture signal decoding method for encoded interlace scanned picture data which has been encoded by an intra coded picture encoding technique or a predictive coded picture encoding technique, the picture data having field intervals, said method comprising predicting a field to be decoded using said predictive coded picture encoding technique by using the two last decoded fields previously decoded by a selected one of said intra coded picture encoding technique or said predictive coded picture encoding technique.

35. A picture signal decoding method wherein the encoded interlace scanned picture data includes head information including ID data to identify the encoding technique that had been used for encoding and further comprising the step of decoding said ID data to distinguish the fields of predicting.

36. A picture signal decoding method for encoded interlace scanned picture data which has been encoded by a bidirectionally-predictive coded picture encoding technique, comprising:

storing picture data from a previously decoded frame and from a following frame to be decoded; and predicting a field in a current frame that has been encoded by using said bidirectionally coded picture encoding technique by using the stored picture data from the previously decoded frame and stored picture data from the following frame but not picture data from another field of the current frame.

37. A picture signal decoding method according to claim 36, wherein the encoded picture data includes header information with ID data to identify which fields in the previous and following frames are to be used to predict said field, and said step of predicting includes decoding said ID data to select the fields which are used for said predicting.

38. A picture signal decoding method according to claim 36, wherein said field is predicted from both fields of said previously decoded frame and from both fields of said following frame.

39. High efficiency encoding apparatus for encoding input picture data in the form of frames, each frame having a pair of fields, comprising:
means for dividing said picture data into macro-blocks, each formed of a two-dimensional array of plural pixels;
motion detection means for detecting frame motion vectors between frames as a function of macro-blocks in respective frames and for detecting field motion, vectors between fields as a function or macro-blocks in respective fields;

first mode selecting means responsive to said motion detection means for selecting a frame prediction mode for carrying out frame motion compensation or a field prediction mode for carrying out field motion compensation;
second mode selecting means responsive to said motion detection means for selecting a frame processing mode for transforming a block of data formed of pixels from said pair of fields or a field processing mode for transforming a block of data formed of pixels from only one field;
predictive encoding means for encoding said input picture data in accordance with said frame prediction mode by using said frame motion compensation or in accordance with said field prediction mode by using said field motion compensation as selected by said first mode selecting means to produce first encoded data, and transform encoding means for encoding said first encoded data by frame or orthogonal transformation or by field orthogonal transformation as selected by said second mode selecting means.

40. High efficiency encoding apparatus for encoding input picture data in the form of frames, each frame having odd and even fields, comprising:
means for dividing said picture data into macro-blocks, each formed of a two-dimensional array of plural pixels;
motion detection means for detecting frame motion vectors between frames as a function of macro-blocks in respective frames and for detecting field motion vectors between fields as a function of macro-blocks in respective fields;

first mode selecting means responsive to said motion detection means for selecting a frame prediction mode for carrying out frame motion compensation or a field prediction mode for carrying out field motion compensation;
second mode selecting means responsive to said motion detection means for selecting a frame processing mode for transforming a block of data formed of pixels from said pair of fields or a field processing mode for transforming a block of data formed of pixels from only one field;
third mode selecting means for selecting a first structure mode or a second structure mode, the first structure mode acting to inhibit encoding or all of the macro-blocks in one frame when said frame processing mode is selected, and the second structure mode acting to inhibit the prediction of an even field of a current frame being encoded from a previously encoded odd field of the same frame;
predictive encoding means for encoding said input picture data in accordance with said selected field or frame motion compensation, said selected frame or field prediction mode and said selected first or second structure mode to produce first encoded data; and transform encoding means for encoding said first encoded data by orthogonal transformation in accordance with the selected processing mode and structure mode.

41. The high efficiency encoding apparatus of claim 40, wherein said prediction mode is operative to produce a bidirectionally predicted picture, and wherein said third mode selecting means is operable to select only said first structure mode for a bidirectionally predicted picture.

42. The high efficiency encoding apparatus of claim 40, wherein said prediction mode is operative to produce a bidirectionally predicted picture and said third mode selecting means inhibits prediction of an even field of a picture from a previously encoded odd field of the same picture for a bidirectionally predicted picture.

43. The high efficiency encoding apparatus of claim 41, wherein said third mode selecting means is operative to inhibit prediction from a previously encoded odd field of a previous frame for a bidirectionally predicted picture.

44. The high efficiency encoding apparatus of claim 40, wherein the frame and field motion vectors include horizontal and vertical components, and said third mode selecting means includes means for determining a parameter indicating the size of motion of an entire picture from a median value of each of the horizontal and vertical components of the motion vector, and said third mode selecting means selects said first or second structure mode as a function of said parameter.

45. The high efficiency encoding apparatus of claim 40, wherein said third mode selecting means includes means for determining correlation between odd and even fields of a frame to select the first or second structure mode depending on said correlation.

46. The high efficiency encoding apparatus of claim 40, wherein said third mode selecting means includes means for determining a squared sum of differences between a macro-block of a current frame being encoded and a similarly positioned macro-block in a previous frame indicated by the motion vector, said third mode selecting means selecting first or second structure mode as a function of the squared sum.

47. The high efficiency encoding apparatus of claim 40, wherein said third mode selecting means includes means for determining correlation between odd and even fields of a frame and means for adding values corresponding to the determined correlation for all macro-blocks present in a current frame being encoded to produce a sum value, whereby said first or second structure mode is selected depending on the sum value.

48. The high efficiency encoding device for encoding picture signals as claimed in claim 40, wherein said structure mode selecting means detects vector anisotropy from the motion vectors of the entire detected macro-blocks and selects said first or second structure mode based on values corresponding to said vector anisotropy summed to values corresponding to correlation between odd and even fields of the current frame being encoded over the entire macro-blocks.

49. High efficiency encoding apparatus for encoding input picture data in the form of frames, each frame having a pair of fields, comprising:
means for dividing said picture data into macro-blocks, each formed of a two-dimensional array of plural pixels;
motion detection means for detecting frame motion vectors between frames as a function of macro-blocks in respective frames and for detecting field motion vectors between fields as a function of macro-block in respective fields;, mode selecting means responsive to the picture data in a macro-block for selecting a frame prediction and a frame processing mode for carrying out frame motion compensation and for transforming a block of data comprised of pixels from both fields in a frame, or a field prediction and a field processing mode for carrying out field motion compensation and for transforming a block of data comprised of pixels from only one field;
predictive encoding means for encoding said input picture data in accordance with said frame prediction mode by using said frame motion compensation and said frame motion vectors or in accordance with said field prediction mode by using said field motion compensation and said field motion vectors to produce first encoded data; and transform encoding means for encoding said first encoded data by orthogonal transformation or by field orthogonal transformation as selected by said mode selecting means.

50. High efficiency decoding apparatus for decoding received encoded picture data comprising;
inverse variable length decoding means for decoding the received encoded picture data to reproduce motion vector information, prediction mode information indicating whether motion compensation is based upon blocks of picture data in which the picture data is derived from one or both fields of a frame, processing mode information indicating whether the encoded data a orthogonally transformed on the basis of a macro-block formed of picture data derived from one or both fields of said frame, and encoded picture data;
inverse transforming means for decoding said encoded picture data by using the inverse orthogonal transform in accordance with said processing mode information to produce first decoded picture data; and prediction decoding means for decoding said first decoded picture data by using motion compensation determined by said motion vector information and said prediction mode information.

51. High efficiency decoding apparatus for decoding received picture data comprising:
inverse variable length decoding means for decoding the received encoded picture data to reproduce motion vector information, prediction mode information indicating whether motion compensation is based upon blocks of picture data in which the picture data is derived from one or both fields of a frame, processing mode information indicating whether the encoded formed is orthogonally transformed on the basis of a macro-block formed of picture data derived from one or both fields of said frame, encoded picture data, and structure mode information indicating if the frame is encoded with a first structure mode in which all of the macro-blocks in one frame are inhibited from being encoded in accordance with the frame processing mode or if the frame is encoded with a second structure mode in which prediction of an even field of a current frame from an odd field thereof is inhibited from all macro-blocks in said frames;

inverse transforming means for decoding said encoded picture data by using the inverse orthogonal transform in accordance with said processing mode information and said structure mode information to produce first decoded picture data, prediction decoding means for decoding said first decoded picture data by using motion compensation determined by said motion vector information, said prediction mode information and said structure mode information.

52. High efficiency decoding apparatus for decoding received encoded picture data comprising:
inverse variable length decoding means for decoding encoded data to reproduce the motion vector information, the prediction mode information indicating which of the block division for motion compensation, the processing mode information indicating which of the block division for orthogonal transform on the basis of a frame in a macro-block or the block division for orthogonal transform on the basis of a field in the macro-block is more efficient and encoded picture data and a macro-block address increment in the header information for the macro-block, address generating means for calculating an address increment value for a frame buffer from said macro-block address increment to find a leading address of each macro-block to accord said leading address to said frame buffer, and motion compensation means for adding a relative address of said macro-block other than said leading address to said frame buffer to access data and for receiving said detected motion vectors, said processing mode information and said structure mode information, said motion compensation means executing prediction between motion-compensated frames of fields in association with said processing mode information and said structure mode information and transmitting the motion-compensated picture signals to said frame buffer.

53. A high efficiency decoding device for decoding picture signals comprising inverse variable length decoding means for decoding encoded data to reproduce the motion vector information, the prediction mode information indicating which of the block division for motion compensation, the processing mode information indicating which of the block division for orthogonal transform on the basis of a frame in a macro-block or the block division for orthogonal transform on the basis of a field in the macro-block is more efficient, encoded picture data and the structure mode indicating the frame encoded with which of a first structure mode of inhibiting the encoding of the entire macro-blocks in one frame in accordance with said the frame processing mode or a second structure mode of inhibiting the prediction of the even field of a current frame being encoded from the odd field thereof for the entire macro-blocks in one frame and a macro-block address increment in the header information for the macro-block, address generation means for calculating an address increment value for a frame buffer from said macro-block address increment to find a leading address of each macro-block to accord said leading address to said frame buffer, and motion compensating means for adding a relative address of said macro-block other than said leading address to said frame buffer to access data and for receiving said detected motion vectors, said processing mode information and said structure mode information, said motion compensation means executing prediction between motion- compensated frames of fields in association with said processing mode information and said structure mode information and transmitting the motion-compensated picture signals to said frame buffer.

54. The high efficiency decoding device for decoding picture signals as claimed in claim 51, wherein said processing information includes, for the bidirectionally predicted picture, the encoding information of selecting the encoding of inhibiting motion compensation and block division for orthogonal transform based on a frame for the entire macro-blocks in one frame and the encoding information of inhibiting the prediction of an even field from an odd field of said picture.

55. The high efficiency decoding deice for decoding picture signals as claimed in claim 54, wherein characterized in that said processing information includes, for the bidirectionally predicted picture, the encoding information of inhibiting prediction from an odd field of a reference frame for forward prediction.

56. A recording medium on which are recorded an encoded bitstream comprising:
encoded picture data produced by using predictive encoding and transform encoding, prediction mode data used in said predictive encoding, motion vector data produced by using said predictive encoding, and processing mode data used in transform encoding.

57. A recording medium according to claim 56, wherein said bistream further comprises structure mode data indicating a first encoding technique in which a second field of a frame is predictable from a first field of the same frame, or a second encoding technique in which said second field cannot be predicted from said first field.